Forthcoming
Photovoltuic
M uteriuIs
Richard H, Bube
Stanford University, USA
Distributed by
World Scientific Publishing Co. Re. Ltd.
P 0 Box 128, Farrer Road, Singapore 912805
USA office: Suite lB, 1060 Main Street, River Edge, NJ 07661
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.
Psalm 19:1,4,6
Interest in photovoltaics has grown rapidly during the last half of the current
century. The importance of alternative energy sources has increased in signif-
icance both for energy supply and ecological conservation reasons. In spite of
limitations due to short-term economic considerations, research and develop-
ment of photovoltaic solar cells has increased and is playing an increasingly
practical role all over the world.
The performance of photovoltaic solar cells is intimately related to the
properties of the materials from which they are made, and many materials
science problems are encountered in the understanding of existing solar cells
and the development of more efficient, less costly and more stable cells.
Every semiconducting material in a suitable electronic environment is ca-
pable of exhibiting properties that might properly be called “photovoltaic” , i.e.
the generation of an electric current and potential difference under absorbed
illumination. What may be considered surprising, therefore, is that so few
materials are known that are able to form photovoltaic devices with sufficient
efficiency to make them of potential interest for practical applications. Allow-
ing for the additional existence of a number of solid solution compounds based
on this small number, it is still noteworthy that only the following materials
have exhibited a solar efficiency in excess of ten percent: silicon, gallium ar-
senide, indium phosphide, cadmium telluride, copper indium diselenide, and
cuprous sulfide. Of these the most versatile is silicon, which can be used to
produce efficient cells in single crystal, polycrystalline or amorphous form.
It is the purpose of this book to describe the properties of these materials
that play an important role in their photovoltaic applications, and to discuss
the experimental and theoretical developments that have led to the leading
contenders among photovoltaic cells for practical applications today. In con-
junction with this is included (where appropriate) a briefer discussion of other
systems that have been tested but have been found to exhibit major prob-
lems primarily in the areas of low efficiency or poor stability under operating
conditions, as well as systems that are considered exploratory today but may
develop into devices of interest in the future. It is not the purpose of the book
to attempt an exhaustive history of the rapidly expanding field of photovoltaic
materials, but to concentrate instead, primarily on developments in the last
decade.
vii
viii Preface
The book starts with a summary of the major properties of the semiconduc-
tor junctions that are involved in conventional photovoltaic devices, together
with the typical quantitative relationships that describe electrical transport in
these junctions. The major purpose of this summary is t o provide a common
framework for the discussions to follow, and it is assumed that the reader has
a basic grounding in semiconductor electronics. It is anticipated that the book
will be used in a teaching or research environment.
The focus of the book is on the materials properties important for photo-
voltaic applications, and does not deal in detail with device design or systems
considerations, except where these are essential to the discussion. It is based
on over twenty-five years of research experience, which has touched a t one
time or another on properties relevant to photovoltaic applications of all of the
major materials.
The book covers research done by a wide variety of investigators in many
countries. I am grateful for the assistance that has been given to me by many
colleagues from all over the world in acquiring reprints of key research papers.
In my own program a t Stanford University, thirty-one students have com-
pleted their Ph.D. research on photovoltaic materials and have added many
significant memories in the process. I gratefully acknowledge support for that
research from the .Air Force Materials Laboratory a t Wright-Patterson Air
Force Base, the NASA Lewis Research Center, the National Science Founda-
tion - Research Aimed at National Needs (NSF-RANN), ERDA, the Division
of Basic Energy Science (BES) of the Department of Energy, the Solar En-
ergy Research Institute (SERI), the Electric Power Research Institute (EPRI),
ARC0 Solar, and the National Renewable Energy Laboratory (NREL). The
research benefited throughout from facilities made available t o Stanford Uni-
versity by the National Science Foundation through the Center for Materials
Research at Stanford University.
I am indebted to the many Visiting Scholars from all over the world who
have participated in this research, and particularly to my immediate colleagues
at Stanford: Dr. Alan L. Fahrenbruch, Dr. David Redfield, and Dr. Adolfo
Lopez-Otero. I am also most thankful for the sun that has shone in my own
life throughout these years because of the constant love and support of my wife
Betty.
Preface vii
Chapter 1. Photovoltaic Materials and Phenomena 1
1.1. Overview 1
1.2. Applications of Photovoltaics 12
1.3. Types of Semiconductor Junctions 16
1.4. More Detailed Photovoltaic Models 25
1.5. Photovoltaic Materials 32
Chapter 2. Crystalline Silicon 34
2.1. Overview 34
2.2. Doping and Lifetime 36
2.3. Major Sources of Recombination 37
2.4. Development of Single Crystal Silicon Solar Cells 41
2.5. Increasing the Efficiency of Silicon Cells 53
2.6. Novel Structures 56
2.7. Thin Multicrystalline Silicon Solar Cells 57
2.8. Microcrystalline Silicon 66
Chapter 3. Amorphous Silicon 68
3.1. Overview 68
3.2. Electronic Structure and Dangling Bond
Defects in a-Si: H 71
3.3. Modes of Defect Generation 74
3.4. Models for Dangling Bond Defects 79
3.5. Overview of a-Si: H Solar Cells 91
3.6. Variations in Deposition Techniques 96
3.7. Material Transport Properties Important
for Solar Cells 97
3.8. Solar Cell Stability 100
3.9. Amorphous Silicon Alloys 106
3.10. Hybrid Solar Cells Involving a-Si : H 109
Chapter 4. Gallium Arsenide and Other 111-V Materials 111
4.1. Overview 111
4.2. GaAs Fabrication Techniques 115
4.3. GaAs-based Solar Cell Development 116
4.4. InP-based Solar Cells 126
4.5. Hybrid Multijunction Cells 131
4.6. Radiation Resistance 132
Chapter 5. Cadmium Telluride and Other 11-VI Materials 135
5.1. Overview 135
ix
x Contents
PHOTOVOLTAIC CELLS
AND PHENOMENA
1.1. Overview
A material (or device) is said to be “photovoltaic” when exposure of the ma-
terial to light that can be absorbed by the material is able to transform the
energy of the light photons into electrical energy in the form of a current and
voltage. The concept is simple and the number of materials that are able to
exhibit photovoltaic characteristics is large. What is not large, however, is the
number of such materials or devices that are able t o make the transformation
of solar radiation to electrical energy with high efficiency, of the order of 20%,
at low cost, and with high stability under operation.
The development of photovoltaic cells can be dated originally to Becquerel’s
discovery in 1839 of a photovoltage produced by the action of light on a n
electrode in an electrolyte solution (Becquerel, 1839). About 40 years later,
Adams and Day observed the photovoltaic effect in selenium (Adams and Day,
1877). Solar efficiencies of about 1% characterized selenium and copper oxide
cells by about 1914. The modern era of semiconductor photovoltaics started
in 1954 when Chapin, Fuller and Pearson obtained a solar efficiency of 6% for
a silicon junction cell (Chapin et al., 1954), a value that was increased t o 14%
by 1958, and to 28% by 1988 (Verlinden et al., 1988). The year 1954 also
dates the announcement of the first all-thin-film cell composed of a Cu,S/CdS
junction (Reynolds et al., 1954) with an efficiency of 6%, later increased to over
9% (Bragagnolo et al., 1980) , which unfortunately had the stability problems
described below. The first mention of a cell based on GaAs was for a 4%
p-n homojunction in 1956 (Jenny et al., 1956); later developments were to
produce cells with efficiencies greater than 30%, as discussed in Chapter 4.
Concern about energy resources motivated a strong surge of interest in solar
cells for terrestrial applications in the early 1970’s. In spite of the fact that the
1980’s and ’90s have been a period in which public and government support for
photovoltaics have been underemphasized, considerable activity and progress
in research and development of solar cells has continued.
1
2 Photovoltaic Materials
ZnSe
.-. ABSORPTION EDGE
I n 1
WAVELENGTH, p m
Fig. 1.1. Average solar spectrum a t the surface of the earth, with the band gaps of various
semiconductors indicated. As discussed in this chapter, a homojunction or a front-wall
heterojunction will absorb all the solar spectrum t o the left of the indicated absorption edge
for a particular material. A back-wall heterojunction will absorb the solar spectrum between
the two band gaps of the materials making up the heterojunction. (Reprinted from R. H.
Bube, "Solar Cells" in Handbook on Semiconductors. Device Physics. Vol. 4C. C. Hilsum,
ed., 1993, p. 797, with kind permission from Elsevier Science - NL, Sara Burgerhartstraat
25, 1055 KV Amsterdam, T h e Netherlands.)
Photovoltaic Cells and Phenomena 3
potential difference between the two ends of the material, 4, which under open-
circuit conditions is known as the open-circuit voltage, &. The polarity of
the open-circuit voltage, therefore, is such as to drive electrons (holes) in the
opposite direction (the forward-bias direction) to that of the electron (hole)
motion in the short-circuit current. Specific examples of photovoltaic systems
are described in more detail in following sections of this chapter.
Fig. 1.2. An early calculation of the theoretical solar efficiency versus semiconductor band
gap for ideal p-n homojunction cells with no surface recombination loss. Curves are shown
for two different junction transport mechanisms: (top) A = 1 for injection dominated current
and (bottom) A = 2 for recombination in the depletion layer. (Reprinted with permission
from J. J. Loferski, J . Appl. Phys. 27, 777 (1956). Copyright 1956, American Institute of
Physics.)
4 Photovoltaic Materials
have material properties such that the photoexcited electrons and holes are able
to be collected by the internal electric field and pass into the external circuit
before they recombine, i.e. a material with a high minority carrier lifetime
and mobility. In order to maximize &, it is also preferred to have the forward
current driven by the photo-induced potential difference be as small as possible
since this current will reduce the potential difference set up by light. The
details of this forward current depend critically on the actual mechanisms of
transport involved, but in general the forward current varies inversely as the
band gap of the material.
From these two purely qualitative considerations, therefore, it may be
concluded that there will be a n optimum band gap or band gap range for
PHOTON ENERGY ( e V )
Fig. 1.3. Optical absorption coefficients of various single-crystal semiconductors commonly
used in photovoltaic devices. (Reprinted with permission from A. L. Fahrenbruch and
R. H. Bube, Fundamentals of Solar Cells: Photovoltaic Solar Energy Conversion, Copy-
right 1983, Academic Press, Orlando, FL.)
Photovoltaic Cells and Phenomena 5
maximizing both J,, and q50c, and hence the efficiency of the photovoltaic de-
vice itself. For interaction with the solar spectrum (see Fig. l.l),this optimum
band gap range lies approximately between 1.2 and 1.8 eV. The results of one
of the first calculations of the dependence of efficiency on band gap for two
different types of junction current are pictured in Fig. 1.2 (Loferski, 1956). De-
tails of the concepts described in the figure caption are discussed later in this
chapter. For comparison, Fig. 1.3 shows the dependence of optical absorption
coefficients on photon energy for various single-crystal semiconductors used in
photovoltaic devices (Fahrenbruch and Bube, 1983).
Fig. 1.4. Simple equivalent circuit for a photovoltaic cell, including a current generator with
total light current I L , a diode with total dark current I D , a series resistance R,, and a parallel
resistance R,. (Reprinted with permission from R. H. Bube, Photoelectronic Properties of
Semiconductors, Copyright 1992, Cambridge University Press.)
make the model even simpler by considering at this point a n “ideal device”
with R, = 0 and R, = co. If we make the simple assumption that the current
generated by light can simply be added to the current flowing in the dark ( “SU-
perposition”), then the current density J flowing in the device in the presence
of photoexcitation can be expressed as
J = J,,[exp(q+/AkT) - 11 - J L (1.1)
Here the first term on the right of Eq. (1.1) is the forward current driven
by the voltage 4, and the second term is the reverse current associated with
photoexcitation. J,, is often called the “reverse saturation current” of the diode,
+
the value of J in the dark for large negative values of in ideal junctions, which
depends on the actual transport mechanism for the diode current (J,, is often
called simply the “pre-exponential coefficient” in practical devices for which
reverse saturation may not occur), and A is the so-called “ideality factor”
that has a value depending on the mechanism of the junction transport (e.g.,
A = 1 if the transport process is diffusion, A z 2 if the transport process is
recombination in the depletion region). Typical variations of total current I
in both the dark and the light as a function of are given in Fig. 1.5. If the
voltage is zero (short-circuit condition), then of course there is zero current in
the dark, but in the light we have
Fig. 1.5. Typical idealized total light and dark current versus voltage curves for a photovoltaic
cell in which the principal cell parameters do not depend on photoexcitation, showing the
open-circuit voltage &, the short-circuit current I,,, and the maximum power point P,.
(Reprinted with permission from A. L. Fahrenbruch and R. H. Bube, Fundamentals of Solar
Cells: Photovoltaic Solar Energy Conversion, Copyright 1983, Academic Press, Orlando,
FL.)
Photovoltaic Eficiency
Since photovoltaic efficiency is of central importance, we see next how this
efficiency is expressed in terms of this simple model. The efficiency of a pho-
tovoltaic solar cell is a maximum when the product of the current density J
and the voltage 4 is a maximum. The efficiency 17 itself can be expressed as
4m = 4oc - ( A k T / q )l n [ ( q h / A k T ) + 11 (1.7)
which can be solved iteratively for 4m. Once a value for 4 , has been obtained,
the value of J,, and hence the maximum power P, = J,&, can be obtained
from Eq. (1.1) by substituting 4 = 4,. The fill factor for a junction describable
by Eq. (1.4) is a function of A and &, increasing with decreasing A and
increasing 4oc (Lindmayer, 1972).
Figure 1.6 gives general insights into the factors that determine the ef-
ficiency in an actual photovoltaic solar cell. The figure shows the various
components to the loss of efficiency in a standard single-crystal silicon cell for
irradiation by sunlight. Specific numbers depend, of course, on the details of
the material and the solar cell, and those given here are intended to be pri-
marily illustrative. The analysis is similar t o that carried out by Wolf (1971).
Starting from the top of the diagram down:
(1) Some of the cell area is obscured by the current-collecting grid, thus re-
ducing the effective cell area: 4% loss;
(2) some of the incident photons are reflected and not absorbed: 2% loss;
(3) some of the photons are absorbed in spurious absorption processes such as
at antireflection coatings, at defects, etc., which do not lead to free carriers:
1%loss;
Photovoltaic Cells and Phenomena 9
G R I D COVERAGE 4.0%
hv < E 18.8%
SPURI(XIS ABSORmION 1 .O% 9
hv > E 29.2%
4
n Q = 0.90 4.5%
f f = 0.78 4.7%
nS 16.m
Fig. 1.6. Power loss chart for a silicon cell showing the percentage of total input power Prad
lost to each of the loss mechanisms. Values of J, = 5 x A cm-2 and A = 1 are assumed.
(Reprinted with permission from A. L. Fahrenbruch and R. H. Bube, Fundamentals of Solar
Cells: Photovoltaic Solar Energy Conversion, Copyright 1983, Academic Press, Orlando,
FL.)
(4) some of the incident photons have a n energy less than the band gap of the
semiconductor (hv < Eg), and hence have insufficient energy to produce
free carriers in the absorbing material: 18.8% loss;
(5) some of the absorbed photons have a n energy larger than the band gap
Eg of the absorbing material; the excess energy (hv - E g ) is non-usefully
dissipated in producing hot carriers which transform the light energy to
heat as the carriers thermalize to near the band edge: 29.2% loss;
(6) not every photoexcited free carrier is collected by the internal electric field;
this is expressed by the quantum efficiency VQ < 1: if V Q = 0.90, this
amounts to a 4.5% loss;
(7) in general the band gap Eg > qq50c, i.e., the energy used to create the
free carriers (hv 2 Eg)is greater than the energy associated with the
open-circuit voltage, and there is therefore an energy loss of 19.2%;
(8) finally in a real solar cell forward-biased t o q5m, with finite R, and R,, the
fill factor f f < 1: if ff = 0.78, this amounts to a 4.7% loss.
10 Photovoltaic Materials
Given the various estimates listed here, the cell efficiency is finally 16.6%.
Losses (1) to (3), and (6) to (8), can be reduced by careful material control
and cell design. Losses (4) and ( 5 ) are difficult to avoid with simple solar cells,
but more complicated structures (graded gap or multijunction structures) can
be designed to help here as well. Photovoltaic cells with efficiency over 30%
have been designed and produced.
(1) Band gap of the absorbing material. The band gap of the absorbing ma-
terial must be small enough to allow absorption of an appreciable portion
of the solar spectrum, and at the same time large enough to minimize the
reverse saturation junction current density J,.
(2) Diffusion length of minority carriers. The diffusion length of minority car-
riers must be as large as possible so that carriers excited by light some
distance from the actual semiconductor junction will be able to diffuse to
the junction and be collected before they recombine with carriers of the
opposite sign. The diffusion length of minority carriers L m i n is given by
L m i n = (Drnin7min)1’2 = [ ( l c T / q ) ( ~ 7 ) , i n ] ~ / ~ , where D m i n is the diffusion
constant, 7,in is the lifetime, and p,in is the mobility, for minority carriers.
It is desired, therefore, to have a material in which the minority carriers
have as large a mobility and lifetime as can be obtained. The value of the
mobility is more or less determined by the choice of material and does not
vary over a wide range. The value of the lifetime, however, is very sensitive
to a variety of phenomena in the bulk and a t the surface that contribute to
recombination of photoexcited carriers. Optimization of efficiency must in-
clude solar cell growth and deposition conditions that maximize the minor-
ity carrier lifetime. Cell design must include photoexcitation of minority
carriers within a diffusion length of the collecting junction.
( 3 ) Desirable junction properties. The actual junction structure and compo-
sition determines the magnitude of the junction transport current density
J , and of the ideality factor A . Formation of the semiconductor junction
must be carefully controlled, therefore, t o produce junctions with as low a
junction current as possible. Various typical mechanisms for the junction
current are summarized in Sec. 1.4.
Photovoltaic Cells and Phenomena 11
(4) In the simple analysis of this section, we have assumed that R, = 0 and
R, = 00 in Fig. 1.4. In real solar cells, however, finite values of these
two resistances will be present and can be a major factor, particularly in
determining the effective value of the fill factor ff. Contributions to the
series resistance R, can arise both from the resistance of the semiconductor
bulk and from the contact resistance to the semiconductor t o complete the
circuit. Problems involving semiconductor doping and control of contact
resistance can play a significant role in some cases. The parallel resistance
R, can be reduced by grain boundaries or other defects that enhance for-
ward junction current and contribute t o an increase in J , and a decrease
in &,=. In polycrystalline thin film solar cells, grain boundaries at the
junction interface can critically affect junction transport properties.
(5) Solar cells are intended for use in exposed areas for long periods of time
without failure. This means that a variety of phenomena that might lead
to a decrease in cell efficiency with time of exposure must be carefully
considered. In some cases, as we shall see, these instability problems may
play a dominant role in determining cell efficiency and utility.
Reference Information
This book provides three kinds of reference information at the end of the vol-
ume: (1) a representative bibliography of books that have been written on the
subject of photovoltaics, with entries listed chronologically; (2) a representa-
tive list of review papers on aspects of photovoltaics, with entries also listed
chronologically; and (3) a traditional list of references to specific research re-
ferred to throughout the book by name of author and date of publication, listed
alphabetically by first author’s name.
In addition, general attention should be called to the published papers pre-
sented at a number of meetings dedicated to the subject of photovoltaics. The
proceedings of the IEEE Photovoltaic Specialists Conference, with meetings
held at 18 month intervals and its 25th meeting in May 1996, are a rich source
of information on progress in the field. At this 25th meeting, for example,
370 papers were presented involving almost 1000 authors. The first World
Conference on Photovoltaic Energy Conversion met in Hawaii as part of the
24th IEEE Photovoltaic Specialists Conference in 1994, and the Second World
12 Photovoltaic Materials
Conference is being held in Vienna in July 1998. There are comparable pro-
ceedings of the Photovoltaic Solar Energy Conferences of the Commission of
the European Communities; the 14th Conference was held in Barcelona in July
1997.
Another valuable source of continuing input is available through the Pro-
gram Review Proceedings and the Annual Reports published through the
years by NSF-RANN (1973-1975), ERDA (1975-1978), and DOE, and the
Solar Energy Research Institute (SERI), now the National Renewable En-
ergy Laboratory (NREL) (since 1978). The Office of Scientific and Tech-
nical Information of DOE also published a bimonthly collection of Current
Abstracts in Photovoltaic Energy. In 1996 this service was replaced by a
listing of abstracts on photovoltaic technology on the World Wide Web at:
http://www.doe.gov/phv/phvhome.html.Beginning in October 1996, it also
became possible to receive a listing of the citations to photovoltaic reports
processed by the Office of Scientific and Technical Information (OSTI) during
the preceding two months, by regular mail, e-mail, or fax.
Some of the background material included in the book, particularly in this
first chapter, is adapted with minor revision from the chapter by the author
on “Solar Cells”, in Handbook o n Semiconductors, Device Physics, Vol. 4, ed.
C. Hilsum, 797-839 (1993), with kind permission from Elsevier Science, NL,
Sara Burgerhartstraat 25, 1055 KV Amsterdam, The Netherlands. Consid-
erable use has also been made of two other books involving the author for
background material and insights into earlier years of research on photovoltaics:
Fundamentals of Solar Cells: Photovoltaic Solar Energy Conversion by
A. L. Fahrenbruch and R. H. Bube, Academic Press, N.Y. (1983), and a chap-
ter on “Photovoltaic Effects” in Photoelectronic Properties of Semiconductors
by R. H. Bube, Cambridge Univ. Press, Cambridge (1992).
problems: air pollution, acid rain, and the greenhouse effect, to name only a
few. In principle nuclear energy might be considered as a way to overcome
these problems, but nuclear fission is a n uninviting prospect in view of the
dangers in plant safety, waste disposal, unfavorable economics, and potential
for misuse, and nuclear fusion is still an uncertain possibility.
The utility of photovoltaics is especially significant with respect to remote
applications, and one of the chief advantages of photovoltaics in the future
is decentralization of the power grid, a consideration also related to remote
energy use.
A Fundamental Limitation
It should be remembered that there is a fundamental thermodynamic con-
straint on our ability to convert energy on earth, considering the accompany-
ing heat, without catastrophic effects. If we were to release on earth, through
the use of either fossil fuels or nuclear energy, heat corresponding to 1% of
the solar power density falling on the earth averaged over a year, the average
temperature of the earth would be increased by about 1"C, with major effects
on climatic conditions all over the world. Today the average level of energy
density consumption in the United States is already about 0.2% of the incident
solar power density, with considerably higher values in urban areas. Only re-
newable energy sources such as solar and wind energy enable us to avoid these
constraints.
lifetime. These three areas are interrelated in a complex way. Already for a
number of years photovoltaics have been applied to a variety of localized tasks,
such as pumping water or telephone communications, in developing nations or
elsewhere in the world where a power grid does not currently exist. It is the
major use of photovoltaics to supply an appreciable fraction of the electrical
power on a general basis that raises critical issues.
highly complex and expensive, but the small amount of material needed
with intense concentration of sunlight again effects a cost reduction.
(3) To be effective as an alternative energy source, a solar cell must generate at
least enough energy in its operating lifetime to pay back both the financial
and energy costs required to produce the cell in the first place, and hope-
fully several times this. It is estimated that an operating lifetime of about
20 years would be a workable value. This lifetime may be determined by
a variety of external factors such as physical damage, corrosion, deteriora-
tion of cell support structures etc., or by what is of more direct concern to
us in this book, by a variety of internal materials-related factors such as
diffusion, photogeneration of defects etc. For cells in space, as compared to
terrestrial-use cells, radiation damage is also a major factor in degrading
performance.
Homojunc tions
A typical energy-band diagram for a homojunction is given in Fig. 1.7. A
homojunction consists of a junction between two portions of the same semi-
conductor, one doped p t y p e and the other doped n-type, hence the name, p-n
junction. Typical details of such an energy-band diagram are shown in the fig-
ure: the vacuum level E,,,, the conduction band edge E,, the Fermi level E F ,
the valence band edge E,, the band gap E c , the electron affinity xs,and the
diffusion potential 4 4 0 . The work function q&v of a semiconductor is defined
as the energy difference (E,,, - E F ) . Since the work function of the p t y p e
portion of the material is greater than that of the n-type portion in Fig. 1.7,
-1
Ec i
\ 1 Eo
XI
0- - 0
Fig. 1.7. Energy-band diagram for a p n homojunction with equal densities of doping in the
p and n-type portions.
18 Photovoltaic Materials
(q4wp- q4wn) = 440, the energy bands between p and n-type portions are
curved, indicating the presence of an internal electric field. Physically one can
think of the diffusion potential q$D as resulting from a transfer of electrons
from n-type to p t y p e material in order to equalize the Fermi energy on both
sides when the junction was formed, giving rise to a positive charge of ionized
donors in an electron-depleted layer near the junction interface on the n-type
side and a negative charge of ionized acceptors in a hole-depleted layer near the
junction interface on the p t y p e side, resulting in a flat Fermi level across the
whole structure. The diagram shown in Fig. 1.7 is specifically for the situation
where the doping of the p and n-type regions is the same, resulting in equal
widths for the depletion layers on both sides of the junction.
Photoexcitation produces free minority carrier electrons in the p t y p e re-
gion, and free minority carrier holes in the n-type region. Each of these then
diffuse toward the junction, and if they reach the junction without being re-
moved by recombination, they pass over the junction, are collected by the
junction field, diffuse through the other portion of the semiconductor and pass
into the external circuit. Since carriers must be excited within by about a
diffusion length of the junction in order to be able t o diffuse to the junction
before recombination occurs, the geometry shown in the insert in Fig. 1.7 is
frequently used, with the portion of the semiconductor on the illuminated side
being much thinner than that on the opposite side. If the semiconductor has
the high optical absorption characteristic of a direct band gap, then the illu-
minated side of the junction must be very thin (= 0.1 pm) t o allow light to
penetrate to within a diffusion length of the junction, but this has the disad-
vantage of causing many carriers to be generated near the surface where the
probability of recombination due to surface defects is greater than that in the
bulk. On the other hand if the semiconductor has a lower optical absorption
characteristic of an indirect band gap, then the portion of the semiconductor
on the opposite side to that illuminated must be quite thick to allow most of
the light to be absorbed, but then much of this absorption occurs more than
a diffusion length away from the junction unless the diffusion length of the
material is large. These considerations show immediately the need for a high-
quality, well-engineered material to serve in an efficient p n homojunction type
of solar cell. High-efficiency, single crystal Si solar cells are usually of the p n
homojunction type.
Heterojunctions
A p n heterojunction is a p n junction formed between two different semi-
conductors with different band gaps and electron affinities. A typical band
Photovoltaic Cells and Phenomena 19
/Front-wall U B a c k - w a l l 1
1 \n
J
P-‘YPe n-‘Y pc
Fig. 1.8. Energy-band diagram for a p n heterojunction with equal densities of doping in
the p and n-type portions, and a choice of material parameters (XI < x 2 , E G <
~ E G 2 ) such
that there are no energy spikes at the junction interface.
diagram for a p-n heterojunction is given in Fig. 1.8, where the p t y p e material
is assumed to have a smaller band gap E G than ~ that of the n-type material
E G ~ As
. in Fig. 1.7, the assumption is made that the doping is of the same
magnitude for both p and n-type materials, giving equal-width depletion layers
on both sides of the junction.
A variety of complex phenomena can occur at the junction interface in such
a heterojunction, which we have simplified in Fig. 1.8, following the Anderson
abrupt-junction model (Anderson, 1962). In this approach we neglect any
effects of interface dipoles or interface states, and recognize that differences
between the electron affinities and band gaps of the two material give rise
to discontinuities A E , in the conduction band and A E , in the valence band,
which can in principle be either positive or negative. Figure 1.8 has been drawn
with the desirable assumption that x1 < ~ 2 so, that an energy spike does not
occur in the conduction band impeding electron transport from the p to the
n-type material.
As indicated in the insert in Fig. 1.8, photoexcitation could be either
on the n-type material (back-wall) or on the p t y p e material (front-wall).
20 Photovoltaic Materials
Back-wall excitation profits from the larger band gap of the n-type material,
which acts essentially like a window even for light that is highly absorbed in the
p t y p e region, and allows the light to penetrate through to the junction with
minimum loss. On the ot,her hand, other problems are introduced which are
related t o the likelihood of lattice mismatch at the junction between the two
semiconductors, and which may in itself produce localized interface states that
facilitate carrier loss through recombination at what is now a kind of “internal
surface”. Such localized interface states may also play a role in increasing the
reverse saturation current density Jo and hence reducing &.
A typical summary of design choices for materials to be used in a hetero-
junction is given in Table 1.1.
Property Criteria
semiconductors (to form an SIS junction) to help reduce the junction currents
that decrease &=.
P+
Fig. 1.9. Energy-band diagram for a p+-p-n buried homojunction or heteroface junction in
which the pf material acts as a large band gap window and a n ohmic contact to the p t y p e
material. Inset shows the typical direction of illumination for use as a solar cell.
22 Photovoltaic Materials
solid solution compound is chosen to provide (a) a large enough band gap to act
as a window layer for solar energy, and (b) a good lattice match to the GaAs,
hence providing a low interface-recombination velocity at the p'-p interface.
Schottky Barriers
The previous three types of photovoltaic junctions all involve junctions between
two semiconductors. A conceptually simpler junction can be obtained from a
Schottky barrier metal contact to a semiconductor (an MS junction).
I I
Semi-
conductor
Metal
n-type semiconductor
Fig. 1.10. Energy-band diagram for a Schottky barrier metal contact t o a n n-type semicon-
ductor, based on the energy parameters of the materials without including interface effects.
p-i-n J u n c t i o n s
What we refer to here as a p i - n junction (see Fig. 3.15 f9r the energy band
diagram of a pi-n junction for a-Si:H) differs from the SIS junction mentioned
above in two major ways: (1) the insulator in an SIS or MIS junction is a
thin layer confined to the junction region, whereas the insulator in a p i - n
junction is usually a thick undoped layer of the same semiconductor where
the principal absorption of light occurs; and (2) transport of carriers in an
SIS or MIS junction occurs by diffusion between the depletion layer and the
contacts, whereas transport in a p i - n junction occurs by drift under a n electric
field that exists throughout the insulator. Since a collecting electric field exists
throughout t,he region where free carriers are being formed by photoexcitation,
there are definite advantages to the p i - n structure. It becomes possible, for
example, to use a semiconductor with desirable properties of other kinds but
with a diffusion length too small to be useful in a conventional p n structure.
One of the most efficient developments in crystalline silicon solar cells is the
p i - n back point-contact (BPC) cell (Swanson et al., 1984) which is discussed
in Sec. 2.4. a-Si:H solar cells are usually fabricated in the form of a p i - n
junction (Carlson, 1977). An example of a heterojunction p i - n structure is
the n-CdS/i-CdTelpZnTe solar cell (Meyers, 1989). Here an n-CdS layer is
deposited on SnOz-coated glass, an i-CdTe layer is deposited by electrodepo-
sition on the CdS, and a p Z n T e layer is deposited on the CdTe.
24 Photovoltaic Materials
Photoelectrochemical Cells
The sixth type of junction goes back all the way to the first discovery of
the photovoltaic effect by observing what happens when a semiconductor-
electrolyte interface is illuminated. Two types of phenomena are of interest:
photoelectrolysis (Bocarsly et al., 1977) and photoelectrochemical cell per-
formance (Gerischer, 1975; Chai et al., 1977; Lewis, 1995). A typical band
diagram for photoelectrolysis is given in Fig. l . l l ( a ) , corresponding t o the
experimental arrangement shown in Fig. l . l l ( b ) , in which electron-hole pairs
are excited in a large band-gap semiconductor, which can then be used to dis-
sociate water. A photoelectrochemical arrangement is shown in Fig. l . l l ( c ) ,
involving an oxidation reaction associated with photoabsorption a t one surface
and a reduction reaction at another surface, providing a flow of electrons t o
the external circuit.
+ “2
I 0, + jH20
H’ //
2-
(b) (C)
Fig. 1.11. (a) Energy-band diagram for photo-electrolysis using a semiconductor; (b) ex-
perimental arrangement for photoelectrolysis using a semiconductor; (c) experimental setup
for a photoelectrochemical cell using a semiconductor. (Reprinted from R. H. Bube, “So-
lar Cells”, in Handbook on Semiconductors. Device Physics. Vol. 4C. C. Hilsum, ed., 1993,
p. 797, with kind permission from Elsevier Science - NL, Sara Burgerhartstraat 25, 1055 KV
Amsterdam, The Netherlands.)
Photovoltaic Cells and Phenomena 25
where the superscript 1 denotes the light condition and other definitions are
similar to those in the dark. The function H ( 4 ) is a voltage-dependent collec-
tion function that describes what fraction of the light-generated carriers are
collected and contribute to the current; we may approximate H ( 4 ) as follows
(Fahrenbruch and Bube, 1974; Mitchell et al., 1977a, b ) .
Collection Function
For the sake of simplicity, consider a p-n heterojunction with a large band-gap
n-type material which transmits photons with energy less than its band gap
without loss to the p t y p e material where they are absorbed. The collection
function H ( 4 ) can be separated into two contributions:
26 Photovoltaic Materials
Here g(4) describes the loss of carriers by recombination in the bulk of the
p t y p e material before they can diffuse to the junction to be collected, and
h ( 4 ) describes the loss of carriers by recombination due to interface states at
the junction interface. Qualitatively one would expect that g(4) will depend
strongly on the photon energy, decreasing for photon energies with lower optical
absorption, since carriers are freed further and further from the collecting
junction and have an increasing probability of recombining while diffusing to
that junction; and that h(4) will be relatively independent of photon energy
but will depend strongly on the electric field a t the interface.
The collection function g(4) can be calculated as follows:
(1.11)
The dependence of g(4) on the photon energy occurs through a(hv). The first
term in Eq. (1.11) describes the complete collection of all carriers created in the
depletion layer with width w, assuming that the drift field there assists in this
collection. The second term expresses the possibility of recombination loss of
such carriers with increasing distance from the depletion layer if the diffusion
length of electrons in the p t y p e material is L,. Integration of Eq. (1.11)gives
the following expression for g(4):
where No+ is the density of ionized donors in the depletion layer. As w(4) > 0 ,
g(4) > 1/P + (1IaLn)l.
A simple expression for the collection function h(4) can be obtained by as-
suming that recombination at the interface is described by an interface recom-
bination velocity S I , and that the recombination probability can be considered
to be a simple competition between crossing the junction without recombina-
tion and recombination a t the interface:
per square centimeter a t the interface with a capture coefficient of PI cm3 s-l,
then S I = NZPI cm s-'. The carrier velocity p E must be considered to have a
maximum value corresponding to the saturation of drift, velocity at high fields.
- 1 .O V Reverse
80
5
I
u
.-
V
E
5
J
40
d
20
0
0.50 0.60 0.70 0.80
Wavelength (pm)
Diffusion
The mode of junction transport corresponding to the smallest values of J, and
to A = 1 corresponds to a diffusion-controlled current over the junction barrier
associated with the injection of electrons [Process (1) in Fig. 1.131 from the
n-type material into the p t y p e material. After injection and diffusion, recom-
bination finally occurs away from the junction in the p t y p e semiconductor
bulk. The current can be expressed as
Photovoltaic Cells and Phenomena 29
ho 2 Recombination
(1.17)
(1.19)
Jo - 4)lw
= qn,[1/(7,07,,)1/21[7TkT/4(4D (1.20)
where w is the depletion layer width, T ~ is , the minimum electron lifetime
when all recombination centers are empty, and r,, is the minimum hole lifetime
when all recombination centers are electron-occupied. The value of A has a
maximum value of 1.8 for symmetrically doped junctions if the levels at which
recombination occurs lie at mid-gap; otherwise values of A between 1 and 2
may correspond to this mechanism (Sah et al., 1957). A plot of In ( J O T p 5 / ' )
versus 1/T has an activation energy of E G , / ~ . For unsymmetrically doped
junctions, the value of A may be equal t o or larger than 2 (Choo, 1968).
with aint= q/kT and A = 1. Two conditions may exist: (1) the thermal
velocity Zlth of the electrons is larger than the interface recombination velocity
s I , so that the current is limited by interfacial recombination, in which case
(1.27)
CRYSTALLINE SILICON
2.1. Overview
Silicon has dominated most semiconductor applications for almost 50 years. It
is the second most abundant material of the earth’s crust, stable and non-toxic;
its technology is highly developed; and in its crystalline form it has an almost
ideal band gap for photovoltaic solar energy conversion. It is not surprising,
therefore, to find that silicon has been the dominant photovoltaic material
in solar cell applications over most of this time period. In 1981 the Si p-n
homojunction cell was the only solar cell widely available commercially.
“Polycrystalline” silicon is the historically generic name for all non-single-
crystalline silicon. In recent years, an effort has developed t o distinguish be-
tween different types of material on the basis of the size of the individual
single-crystal grains in the material, and to reserve the specific name “poly-
crystalline” for just one of them (Basore, 1994). A complete “family tree” for
silicon can therefore be listed as follows: (1) single crystal - just one grain
(> 10 cm) characterizes the whole material; (2) multicrystalline (mc-Si) -
there are a number of different grains in the material but the grain size is
relatively large, of the order of 1 mm to 10 cm; (3) polycrystalline - many
grains are present in the material with grain dimensions of the order of 1 pm
to 1 mm, as is also the case for polycrystalline films of CdTe (Chapter 5 ) or
CuInSez (Chapter 6); (4) microcrystalline (pc-Si) - single crystal regions in
grains can still be identified in x-ray diffraction patterns, but the grain size is
< 1 pm; (5) amorphous (a-Si) - no single-crystal regions identifiable by x-ray
diffraction occur since no long-range order is present. We discuss the varieties
of crystalline silicon in this chapter and amorphous silicon in Chapter 3.
The cost of single-crystal Si solar cells is greater than that of its closest
competitors, thin-film multicrystalline Si cells, and the impetus is often present
to trade the lower cost of the multicrystalline cells for the higher efficiency of
the single crystal cells. Single-crystal Si solar cells require very high material
quality; surfaces must be effectively passivated to reduce recombination there,
and the bulk properties must also be of high quality because of the long optical
penetration distances associated with the indirect band gap of Si. Specially
34
Crystalline Silicon 35
Efficiency (Yo)
30
20
10
0
1940 1950 1960 1970 1980 1990 21 DO
Fig. 2.1. The increase in efficiency of single-crystal silicon laboratory solar cells. (Reprinted
with permission from M. A . Green, Szlicon Solar Cells: Advanced Principles & Practice,
Centre for Photovoltaic Devices and Systems, Univ. of New South Wales, Sydney, 1995.)
36 Photovoltaic Materials
Phosphorus and boron are the most widely used donor and acceptor dopants in
single-crystal Si. As the density of dopants is increased in the active absorbing
Table 2.1.
*The abbreviation UNSW stands for University of New South Wales, Australia.
Crystalline Silicon 37
layers of solar cells, two effects occur that degrade the electronic properties of
Si.
The lifetime is decreased both because of an increase in Shockley-Read type
recombination (electron-hole recombination through imperfections, giving up
the recombination energy as photons or phonons), and by the onset of Auger-
type recombination (in which the recombination energy is given up t o another
free carrier - see Sec. 2.4) for impurity densities in excess of 5 x 1017 cm-3
(Fischer and Pschunder, 1975; Redfield, 1978, 1979, 1980).
The carrier mobility, and hence the diffusion coefficient, is decreased be-
cause of an increase in charged impurity scattering, noticeable for impurity
densities greater than 10l6 ~ m - ~Accurate
. measurements of the mobility of
minority-carrier holes in phosphorus-doped silicon with doping in the lo1’
cm-3 range have been made using independent measurements of diffusion
length and minority-carrier lifetime t o obtain minority-carrier mobility values
(del Alamo et al., 1985; Swirhun et al., 1986; del Alamo et al., 1 9 8 7 ~ ) .Since
the absorber layer in the solar cell generally has doping of less than 10l6 ~ m - ~ ,
the mobility effect is usually negligible.
Deep-level impurities in Si may also decrease the lifetime of carriers through
recombination. Such impurities in roughly decreasing order of effect include
Ti, Zr, V, Na, Au, Cu, Fe, Mg, Cr , Mn and Ni, with tolerable densities varying
from 4 x 1013 cmP3 for T i t o 4 x 10l6 cm-3 for Ni (Hill et al., 1976; Davis
et al., 1980; Shimura, 1994). Particularly harmful are Na, Cu, and Fe since
they also have high diffusion coefficients in Si.
Introduction of dopants into the Si is usually by the process of diffusion,
which may be done in a variety of ways, often leading to a high concentration of
the dopant near the surface. This highly-doped layer has a decreased lifetime
and must be avoided or later removed in order not to have a “dead-layer” a t
the surface.
Bulk Recombination
As described in Sec. 2.2, a major source of bulk recombination is associated
with imperfections in the bulk of the Si. The desired result is to minimize
recombination throughout the bulk through the use of high-quality pure Si
38 Photovoltaic Materials
HCI, TCE
Contaminants
I Ep1 s1 I
I
Denuded
Zone . .
*. Intrinsic Gettering
.. .* dcfcclr
Silicon
p s s ,D e y
111 P dilluslon
poly-SI lilm Si,N, lilm Ion Implantation
f 1 Laser irradialion
Sand-blasling
Mechanical abrasion
I v J
I Extrinsic Gettering I
Fig. 2.2. Schematic illustration showing various gettering techniques for a silicon wafer
(Reprinted with permission from F. Shimura, Semiconductor Silicon CnJstal Technology,
Copyright 1989, Academic Press, Orlando, FL.)
starting material, thinning the bulk quasi-neutral regions, and limiting the
doping t o the minimum levels necessary t o achieve cell performance (Green,
1984).
This direct avoidance of impurities sometimes proves either too difficult or
too expensive, and other methods referred to as “gettering” are used to reduce
the impurity density in the bulk (Green, 1995a). These are shown schemati-
cally in Fig. 2.2 (Shimura, 1989). (a) A standard method in microelectronics
is “intrinsic gettering” which involves the formation of SiOz precipitates in the
bulk that act as gettering sites for further processing (Sardana, 1985) after de-
pletion of oxygen from the surface by out-diffusion at high temperatures. Since
this method results in very high quality material near the surface, but poor
quality material in the bulk, it is not suitable for photovoltaics. (b) A second
Crystalline Silicon 39
Surface Recombination
Non-contacted areas of the surface of the Si provide locations where recombi-
nation is enhanced. This recombination is generally minimized by the growth
of a high-quality thermal oxide. The thickness of the oxide (14 nm to 80 nm)
does not appear to be an important parameter (Kopp e t al., 1992).
The magnitude of the effective surface recombination velocity Sefffor pho-
toinduced carriers at the Si-Si02 interface as a function of injection level has
been investigated (Aberle et al., 1992; Glunz et al., 1994a, 199413). A strong de-
pendence of Seffat the Si-SiOa interface on injection level and doping concen-
tration was found as shown in Fig. 2.3, which may be consistently interpreted
in the following way.
Results indicate that the electron capture cross-section of interface states is
much larger than their hole capture cross-section (Aberle et al., 1992). Because
of this difference in cross-sections, which makes the interface states behave sim-
ilarly to donors with a positive charge when unoccupied, the recombination
properties are quite different for n-type and p t y p e material. For n-type sur-
faces, there is a small density of holes near the surface, and hence the capture
of holes is the rate-determining recombination process; the net recombination
rate is small because of the small hole density and the small hole capture
cross-section. For p t y p e surfaces, however, the situation is more complex and
changes with increasing output voltage of the cell. At low voltages capture
of electrons is the rate-determining step, and the recombination process is
more effective than in p t y p e material because of the larger electron capture
cross-section. With increasing voltage across the cell, however, the situation
is reached where the hole capture becomes the rate-determining step, and the
effective surface recombination decreases, eventually t o values comparable to
40 Photovoltaic Materials
1000,. I . I I I Ir I I
-< 100-
-
b,.
a 10 R c m p-type
E
0
W
-
~ 102...._.._._.._,.......
”’
N, =1.4x1Ol6 cm-’ -
- N, =7.5x101’ cm-’
-N, =6.5x101’ cm-’
.... ... .... N, =1.Ox1O1’c m - ’
I I I
1
Fig. 2.3. Measured injection dependence of the effective surface recombination velocity,
S,R, for differently doped silicon. (Reprinted with permission from S. W. Glunz et al.,
f Z t h European Photovoltaic Solar Energy Conference, p. 492 (1994). Copyright 1994, H. S.
Stephens and Associates, Bedford, United Kingdom.)
those found for n-type material. It must be kept in mind, however, that
specific fabrication details may be significant in determining the details of
surface recombination.
Recent developments in techniques for surface and bulk passivation are re-
ported by Loelgen (1995), who describes the effects of doped layers at the sur-
face on reducing the surface recombination, and of Al-gettering t o improve the
silicon bulk quality (Loelgen et al., 1994a,b);and by Leguijt (1995), who inves-
tigated the effects of surface passivation by low temperature plasma enhanced
chemical vapor deposition (PECVD) of silicon oxide or silicon nitride. Simi-
larly Rosmeulen et al. (1995) reported a 19.5% efficient solar cell passivated
by oxides deposited by low pressure chemical vapor deposition (LPCVD), and
Demesmaeker et al. (1994) reported promising results with low-temperature
surface passivation by means of a Tetra-Methyl-Cyclo-Tetra-Siloxane (TM-
CTS) LPCVD oxide.
Crystalline Silicon 41
(4
top metal finger antireflection coaling
/
n-type
/
rear metal
P-tYPe
IL
Fig. 2.4. (a) Form of the first modern silicon cell (Chapin et al. 1954); (b) standard design
for a space silicon cell (Mandelkorn et a!. 1962; Smith et al. 1963); (c) the shallow junction
“violet cell” (Lindrnayer and Allison, 1973). (Reprinted with permission from M. A. Green,
Szlzcon Solar Cells: Advanced Prznczples 0 Practzce, Centre for Photovoltaic Devices and
Systems, Univ. of New South Wales, Sydney, 1995.)
Crystalline Silicon 43
Fig. 2.5. Energy band diagram for a typical Si p+-n-nf solar cell with the n-layer thickness
exaggerated, and including a back surface field (BSF) created by A1 diffusion. (Reprinted
with permission from A. L. Fahrenbruch and R. H. Bube, Fundamentals of Solar Cells:
Photovoltaic Solar Energy Conversion, Copyright 1983, Academic Press, Orlando, FL.)
principal junction transport mechanism (see Eqs. 1.17 and 1.18). It was this
reasoning, leading to heavy doping of the n-type layer, that was responsible
for the long period of fairly flat solar cell efficiencies shown in Fig. 2.1. These
earlier cells had almost the same short-circuit current (= 40 rnA/cm2) as the
best modern cells, but an appreciably smaller open-circuit voltage (0.59 V
compared to 0.71 V). It is based on the more general version of Eq. (1.18) for
the reverse saturation current density J,, for uniform doping on both sides of
the junction,
The C N R Cell
Another early type of Si p-n junction solar cell was the T o m s a t non-reflective’’
(CNR) cell (Haynos et al., 1974; Arndt, 1975; Rittner and Arndt, 1976), some-
times also called the “black cell”. It was characterized by a physically tex-
turized front surface designed to reduce reflection losses and to cause photo-
generation closer to the junction by allowing light to travel in paths that are
not normal to the junction interface. Such cells typically had q50c = 0.59 V,
J,, = 46 mA/cm2, f f = 0.77, J , = 6 x A/cm2, A = 1, = 15.5%
Crystalline Silicon 45
for AMO, and Q = 18% for AM1 radiation. General issues related to coupling
sunlight to solar cells in a useful way are discussed by Luque (1993).
I
hv
Fig. 2.6. A cross-section of a region in a Si back-point contact solar cell near one of the A1
bussbars. (Reprinted with permission from R. A. Sinton et al., I E E E Electron Device Lett.
EDL 7,567 (1986). Copyright 1986, IEEE, NY.)
alternate p t y p e and n-type small point contacts only on the back surface. The
small size of the contacts allows a considerable decrease in recombination, but
contact resistance must be minimized. Careful attention is paid t o minimizing
the surface recombination velocity over the entire surface of the cell, and to
maintaining maximum bulk carrier lifetimes since carriers photoexcited near
the top of the cell must diffuse t o the contacts a t the back.
These were the first cells to exceed 22% efficiency under normal terrestrial
illumination. The most efficient cell reported has a thickness of 120 pm, an
area of 0.15 cm2 with a metal-finger length of 3 mm, and an efficiency of 28%
at 15 W/cm2 of AM1.5 illumination (about 150 suns). Figure 2.7 shows how
the efficiency varies with the concentration of the illumination, and Fig. 2.8
shows the calculated deconvolution of the recombination current a t the maxi-
mum power point for the cell of Fig. 2.7 (Sinton et al., 1986b). Over this range
of concentration, the measured open-circuit voltage increases from 650 mV a t
1 sun to 840 mV at 600 suns. The importance of recombination in the periph-
eral region, which dominates a t low concentrations, decreases with increasing
concentration, as does the recombination a t surface and bulk defects in the il-
luminated regions of the cell. Recombination in the emitter (diffused) regions
Crystalline Silicon 47
28 - 24*1 t
8 4 f++I
5 26-
f f
c
.-a
-
.?
c
w
24 -
-
I
$I :
22'L
ii-
!
?
)
z
00
Auger
j 60- perimeter
0
c emitter
.G
4-
40-
a surface
.-
t
I
0
V
U
1 10 100 1000
Concentration, s u n s
Fig. 2.8. Calculated deconvolution of the recombination current at the maximum power point
of the cell of Fig. 2.7 into its components as a function of the concentration value. (Reprinted
with permission from R. A. Sinton et al., Copyright 1986, Sandia Report SAND86-0058/1,
Albuquerque, New Mexico.)
finger
\
"iyerted" pyramids
thickness about 30 times that of the physical thickness of the cell. Typical
output values for PERL cells are J,, = 41 mA/cm2, &c = 710 mV, ff = 0.83
and 77 = 24%. An efficiency greater than 45% has been measured with this
structure for 1.02 pm monochromatic light (Green et al., 1992) and of 46.3%
for 1.04 pm wavelength light (Zhao et al., 1995).
As far as recombination losses in these cells are concerned, it is believed
that bulk recombination can be made negligible using present designs with
appropriate processing conditions (Green, 1993a). Rear surface recombination
in non-contacted areas is not important at open circuit (the effective recom-
bination velocity decreases from large values of lo4 cm/s t o low values below
30 cm/s as the voltage increases from short-circuit to open-circuit values), but
is important at the maximum power point voltage, resulting in a reduction in
fill factor. At the open-circuit voltage, recombination in the top diffused layer
and its contact region dominates the behavior. A detailed investigation of the
limiting loss mechanisms in PERL cells has been made by Aberle et al. (1995).
22 Oxide elchinp
Fig. 2.10. Illustration of a processing scheme for a high-efficiency silicon solar cell with a
boron back surface field. (Reprinted with permission from J. Knobloch et a!., 23rd ZEEE
Photovoltaic Specialists Conference, p. 271 (1993). Copyright 1993, IEEE, NY.)
c
5 *
L
L
3
0
0
0.0 0.1 02 0.3 0.4 05 0.6 0.7
VoItoge p]
Fig. 2.11. Current-voltage characteristics for the best solar cells from FZ-, CZ- and multicrys-
talline silicon prepared by the processing schedule of Fig. 2.10. (Reprinted with permission
from J. Knobloch et al., 23rd I E E E Photovoltaic Specialists Conference, p. 271 (1993).
Copyright 1993, IEEE, NY.)
52 Photovoltaic Materials
350
EJ)
C
3 250
c
.g 200
zb' 150
s
'
'jo 100
aJ
50
o ~ " " ' " " " " ' ' ' ' " '
0 200 400 600 800 1000
Dislocation Density (#/cmZ)
Fig. 2.12. Dependence of effective diffusion length on the dislocation density in CZ silicon
solar cells. (Reprinted with permission from K. L. Pads et al., 23rd IEEE Photovoltaic
Specialists Conference, p. 209 (1993). Copyright 1993, IEEE, NY.)
(Werner et al., 199513). For a terrestrial cell, the predicted theoretical maxi-
mum efficiency including this process is 44.2%.
Another way to increase the solar cell current is to make use of multiple-step
absorption through defect levels in the forbidden gap in silicon. The danger
here, of course, is that an increase in defect density corresponds to an increase
in recombination loss. It has been concluded, however, that impurity levels
lying closer to the band edge than to midgap can in principle improve the
current output without major loss in cell voltage.
These various methods of increasing the current in a Si cell by the effective
use of subgap absorption emphasize the need for an accurate knowledge of the
absorption of intrinsic silicon over the whole solar spectral range. Keevers and
Green (1995) used sensitive subgap photocurrent measurements on high effi-
ciency PERL silicon solar cells to determine the absorption coefficient between
1.19 and 1.45 pm. These measurements have been augmented by previous data
for shorter wavelength light and the absorption over the whole range from 0.25
to 1.45 pm is listed by Green (1995a). A plot of the spectral dependence of
the absorption coefficient corresponding to these data is given in Fig. 2.13.
They cover 14 orders of magnitude of the absorption coefficient, and the long
wavelength fine structure giving rise to inflection points at 1.18 pm, 1.25 pm,
and 1.34 pm corresponds to the l-phonon edge, the 2-phonon edge, and the
3-phonon edge respectively.
Silicon Spheres
Solar cells have been fabricated using inexpensive material as feed stock, and
then growing single crystal 1-mm-diameter spheres of silicon by melting silicon
particles (Levine et al. 1991; Levine, 1992; Ahrenkiel et al., 1996). The process
in which small-diameter spheres are grown provides a purification step by di-
rectional solidification, allowing for the use of highly impure starting material.
The starting material is melted and then single crystal spheres are formed by
the surface tension from small droplets of melted silicon. Impurities segregate
to the surface of the spheres, which is removed in a repeated process to produce
sufficient purification for solar cell use. The spheres are commonly B-doped
p t y p e , and a p n junction is formed on the entire outer surface by P diffusion.
The spheres are pressed into small orifices in an A1 foil that makes contact
Crystalline Silicon 57
with the n-type surface, the top of the spheres are etched back t o reveal the
D-type interior doping, and a contact is made to the inner p t y p e material
with a second foil, the two foils being electrically insulated from each other by
deposition of a polymer. Because of their geometry the cells are particularly
2fficient a t collecting diffuse light. A power module consisting of forty 100 cm2
:ells has demonstrated an efficiency of 10.3%.
Recombination lifetime studies on these silicon spheres has been carried
mt using radio-frequency photoconductivity decay (Ahrenkiel et al. , 1996),
-omparing cells made from metallurgical grade silicon with those made from
4ectronic grade silicon. After two melting cycles, the lifetimes of cells made
From electronic grade material are a n order of magnitude larger than those
made from metallurgical grade material.
Superlattices
The possible application of superlattice design to silicon solar cell developments
should be included in this discussion. To date most attention has been given
to silicon/germanium superlattices, in which the presence of strain introduces
new effects of possible ultimate interest (Barnham et al., 1990, 1991).
General
The use of thin multicrystalline silicon cells offers five major advantages: (1)
less material is needed without loss of current provided that light-trapping
is adequate; (2) somewhat lower quality material can be used since smaller
diffusion lengths can be tolerated in thin cells; (3) theoretically thin cells can
produce higher open circuit voltages and hence efficiencies than thick cells with
the same bulk diffusion length; (4) thin cells are more radiation-resistant; and
( 5 ) a major advantage in economy of growth.
The use of multicrystalline materials in solar cells, however, introduces
several new challenges. Major considerations concern the control of nucle-
ation and growth of crystalline Si on foreign substrates at low temperatures,
the need for a high-quality light-trapping scheme since the thickness of the
58 Photovoltaic Materials
Fig. 2.14. Energy band diagram for a grain boundary in an n-type polycrystalline material.
decrease the height of the grain boundary potential barriers, thus increasing
the majority carrier mobility, and the film conductivity (Bube, 1992).
A general schematic diagram for a thin film polycrystalline solar cell is given
in Fig. 2.15, illustrating the difference between horizontal grain boundaries ly-
ing within n- or p t y p e material, and vertical grain boundaries crossing the
interfaces between n- and p t y p e regions. Horizontal grain boundaries may re-
duce the short-circuit current because of a decrease in minority carrier lifetime
through an increase in the density of sites for recombination, and vertical grain
boundaries may reduce both the short-circuit current and the open-circuit volt-
age if they intersect the junction boundary and provide effective shorting paths
that increase the reverse saturation current. In addition, grain boundaries may
also increase the diffusion of impurities by rapid diffusion along grain bound-
aries, and provide favorable sites for impurity segregation due to a kind of
gettering effect.
It is believed that grain boundaries in Si are symmetrically depleted with
about the same activity in n-type and p t y p e material , and that the elec-
tronic properties of these grain boundaries are not dominated by such intrinsic
properties as dangling bonds, but by extrinsic (impurity-related) properties
(Werner et al., 1 9 9 4 ~ ) .The presence of oxygen and carbon, as well as many
transition metals within the boundary plane, strongly alters the properties of
Si grain boundaries.
The formation of photoinduced metastable defects associated with dan-
gling bonds has been reported for hydrogenated polycrystalline silicon (Nickel
et al., 1993; Redfield and Bube, 1996) with many similarities to the effects ob-
served in amorphous silicon as described in Chapter 3. Electron spin resonance
60 Photovoltaic Materials
depletio
region
Fig. 2.15. Thin film polycrystalline silicon solar cell showing grain boundaries: (a) horizontal
boundary in bulk region; (b) horizontal boundary in depletion region; (c) vertical grain
boundary. (Reprinted with permission from M. A. Green, Silicon Solar Cells: Advanced
Principles €9 Practice, Centre for Photovoltaic Devices and Systems, Univ. of New South
Wales, Sydney, 1995.)
(‘Ribbod7 Technology
Various procedures belonging to variations on a “ribbon” technology in which
sheets/ribbons/foils of Si for solar cells are produced, have been developed
to allow lower cost of multicrystalline Si solar cells, while giving up some.of
the efficiency (Goetzberger and Rauber, 1988). They include (1) dendritic
web (WEB) with shaping by two dendrites (Hopkins et al., 1987), (2) edge-
defined film-fed growth (EFG) with shaping by a graphite die (Wald, 1987),
(3) melt-spinning (MS) with casting in a graphite die (Madea et al., 1987),
(4) ramp-assisted foil technique (RAFT) with foil growth on a reusable ramp
(Beck et al., 1987), (5) horizontal supported web (HSW) with wedge-shaped
62 Photovoltaic Materials
Passivation procedures
The grain boundary potential barriers in both n- and p t y p e silicon are re-
duced by the diffusion of monatomic hydrogen into the grain boundaries (Au-
couturier, 1991; Sopori et al., 1991; Seager et al., 1982; Rohatgi et al., 1993).
There is a decrease in the density of trapping sites for majority carriers, and a
decrease in the recombination velocity by a couple of orders of magnitude for
minority carriers. Still the fundamental microscopic mechanism is not com-
pletely understood. The extent of the passivation depends on the specific ori-
gin of the boundaries, and results not only from removing dangling bonds, but
also electronic defects related to impurity segregation. Both grain boundaries
and dislocations in LPE-grown Si appear to have a much lower recombination
velocity than those in the substrate, suggesting intrinsic gettering and bond
reconstruction during the near-equilibrium LPE growth of the film.
Another possibility for effective passivation of thin-film solar cells is the
plasma enhanced chemical vapor deposition (PECVD) of silicon nitride (Mich-
iels et al., 1990; Szlufcik et al., 1994; Schetter et al., 1995). It can serve as a n
antireflection coating and passivating layer, and can be deposited in existing
commercial systems. Strong enhancement of the bulk and surface passivation
from a plasma nitride layer has been reported during a fast firing for screen
printing metallization.
Gettering techniques
Effective gettering for a thin-film solar cell where the entire volume is electron-
ically active, requires the movement of impurities to the surface. Techniques
with P and A1 gettering have been reviewed (Verhoef et al., 1990; Gee, 1991;
Rohatgi et al., 1993; Porre et al., 1993; Perichaud et al., 1993), as has getter-
ing with P and C1 (Jastrzebski 1995). Verhoef et al. (1990) reported diffusion
experiments indicating that impurities such as copper, gold and nickel are
effectively gettered at the aluminum-doped layer. Gettering in single- and
multicrystalline silicon has been compared, using P, Al, and C1 gettering, as
well as back-side hydrogen gettering, and ultrasound treatments (Sopori et al.,
1996).
Crystalline Silicon 65
shift, and in order to reach a high spatial resolution, the infrared laser beam
is focused on the sample (light-spot diameter < 100 pm).
Subsequent work raised the efficiency t o 4.6% and indicated that exposure t o
solar illumination did not degrade the totally pc-Si:H cells. An a-Si:H/pc-
Si:H stacked cell showed an efficiency of 9.1%, which did not degrade after
280 hours of intensive illumination with a sodium lamp with brightness of
about six suns. It is suggested that the electronic behavior of the pc-Si:H is
controlled by the crystalline phase in these materials. A beginning has also
been made on understanding the transport properties of compensated pc-Si:H
(Wyrsch et al., 1996) and the enhanced optical absorption of pc-Si:H (Beck
et al., 1996). The potential utility of pc-Si(l--s)Cz interface layers in a-Si:H
and mc-Si solar cells has also been investigated (Ma et al., 199413).
Microcrystalline-crystalline Si heterojunction solar cells were made using
a thin p + pc-Si:H window layer on top of 1 ohm-cm n-type single-crystal Si
(van Cleef et al., 1996). These cells had the following structure: Ag/ITO/p+
pc-Si:H/a-Si:H/n-Si/Al. A buffer layer consisting of intrinsic a-Si:H deposited
at low temperature was found to be essential in achieving high open-circuit
voltage and efficiency (12.2%).
CHAPTER 3
AMORPHOUS SILICON
3.1. Overview
Amorphous silicon is a thin-film material without long range atomic order that
can be deposited by a variety of techniques suitable for large area production.
In this form, however, amorphous silicon has a high density (of the order of
1019 cmP3) of coordination defects (“dangling bonds”) corresponding to de-
partures from the local tetrahedral coordination between four silicon atoms
(see Sec. 3.2). They act as recombination centers t o greatly reduce the carrier
lifetime and the carrier diffusion and drift lengths, and t o pin the Fermi energy
in such a way that the material cannot be effectively doped n- or p t y p e and
is not electronically useful. It was discovered, however, that the incorpora-
tion of about 10% hydrogen (from glow-discharge decomposition of silane gas)
during the deposition process greatly reduces the density of these defects to
about 10l6 cm-3 (Chittick et al., 1969), and this hydrogenated material can
then be doped n- or p t y p e to make useful semiconductor devices (Spear and
LeComber, 1976). A promising new thin-film silicon alloy, a-Si:H, had been
discovered.
Perhaps of greatest significance for photovoltaic applications, the hydro-
genated amorphous silicon alloy (a-Si:H) also has a higher-energy absorption
edge and a larger optical absorption constant for solar radiation than crys-
talline silicon. a-Si:H has a well-defined optical threshold at about 1.75 eV,
considerably larger and sharper than the indirect band gap of 1.1 eV of crys-
talline Si, as shown in Fig. 3.1, and only 1 to 2 pm thickness of a-Si:H is
required to absorb virtually all of the light above the absorption edge.
In the 20 years since this initial work, a voluminous literature has grown
up related to amorphous silicon. As examples we cite a general discussion of
amorphous silicon in a four-volume series edited by Pankove (1984), a two-
volume set edited by Fritzsche (1989), a book by Street (1991), and a two-
volume series edited by Neber-Aeschbacher (1995) containing thirty-six review
papers and sixteen original papers on many aspects of hydrogenated amorphous
silicon and its alloys. A portion of a book by Redfield and Bube (1996) is
68
Amorphous Silicon 69
Fig. 3.1. A comparison between the optical absorption coefficient for single crystal Si and a-
Si:H a s a function of wavelength. Long-wavelength d a t a are from Loveland et al. (1973/74).
(Reprinted with permission from D. E. Carlson and C. R. Wronski, A p p l . Phys. Lett. 28,
671 (1976). Copyright 1976, American Institute of Physics.)
concerned particularly with the properties of the defects and their induction
by photoexcitation.
It appears, therefore, that a-Si:H should be a n excellent choice for thin-film
solar cells: It is related to the general silicon technology and it can be relatively
inexpensively deposited in thin film form with high uniformity over large areas
on inexpensive substrates (e.g. glass) at low temperatures. Problems are asso-
ciated with the high density of defects still present even in the hydrogenated
material, and with the observation that the density of these defects increases
upon a variety of treatments of the material (see Sec. 3.3), particularly t o
exposure to high-intensity light (optical degradation). In spite of these insta-
bilities, a-Si:H has often been considered the front-runner in competition with
polycrystalline materials (see Chapters 5 and 6 ) for thin-film solar cells.
70 Photovoltaic Materials
UNDOPED
a-Si
- hu
-Al
p/i/n JUNCTION
)TOJ
UNDOPED
0-Si
J UNDOPED -
o Si
d CONTACT PAD
hu
Pt SCHOTTKY
BARRIER
t t
n' a - S i ZroZ
Fig. 3.2. Three types of a-Si:H junctions: (top) a heterojunction between SnO2 and
a-Si:H, (center) a p z - n junction with p + - and n+-a-Si:H layers with undoped a-Si:H between
them, and (bottom) a Schottky barrier between Pt and n+-a-Si:H doped with P. (Reprinted
with permission from D. E. Carlson, IEEE Trans. Electron Devices ED-24, 449 (1977).
Copyright 1977, IEEE, NY.)
A variety of junction structures has been tested for solar cells using
a-Si:H, as indicated in Fig. 3.2. The most effective is generally agreed to be
a p a - n structure, consisting of an undoped insulating region between doped
p and n-type end regions, which makes it possible for the main absorber
portion of the cell to have the longer lifetimes of undoped a-Si:H, and for
the carrier collection process to be dominated by drift rather than diffusion.
Initial efficiencies as high as 12% had been obtained for single-junction cells
by the late '80's (Wronski, 1986; Hubbard, 1989; Hubbard and Cook, 1989).
The general improvements in a-Si:H solar cell efficiency are summarized in
Amorphous Silicon 71
0
1975 1980 1985 1990
Year
Fig. 3.3. Improvement of a-Si:H based solar cells from 1975 to 1989. (Reprinted with per-
mission from E. S. Sabisky and J. L. Stone, 20th IEEE Photovoltaic Specialists Conference,
p. 39 (1988). Copyright 1988, IEEE, NY.)
Fig. 3.3 with data shown for small area single and stacked cells and for large
area modules (Sabisky and Stone, 1988). It is projected that the efficiency of
a multijunction cell may reach 20% by the year 2000 (Maruyama et al., 1995).
It is the effects of optical degradation that have limited the practical effi-
ciency of a-Si:H solar cells, and many attempts have been made to determine
the exact mechanisms involved in this process, to overcome defect formation
itself, or to redesign the solar cells so that the optical degradation effects are
minimized. These effects are discussed further in Secs. 3.3 and 3.8.
102’ -l
Fig. 3.4. Density of electron states in the band gap of a-Si:H, characterized by exponentially
varying valence and conduction band tails, and states associated with dangling bond defects
near the middle of the gap. (Reprinted with permission from D. E. Carlson and S. Wagner,
in Renewable Energy: Sources for Fuels and Electricity, eds. T . B. Johansson et al., p. 403,
Copyright 1993, Island Press, Washington, D.C.).
Amorphous Silicon 73
mobility edge exists in the energy spectrum of mobility. The gap that does
exist, between the mobility edge for electrons in the conduction band and the
edge for holes in the valence band, is a mobility gap, which is quantitatively
similar to the optical absorption threshold. Figure 3.4 gives a schematic plot
of the density of defect states in a-Si:H as a function of electron energy above
the valence band (Carlson and Wagner, 1993).
Diffraction patterns of quality a-Si:H show that nearly every Si atom is
bonded to four others in the same tetrahedral coordination as in crystalline
Si, with first and second nearest-neighbor distances similar to the values in
crystalline Si. The major difference is that in the amorphous material there is
an approximately 10" range of bond angles centered at the tetrahedral value
of approximately 109". The densities of amorphous and crystalline Si are
also approximately the same. The ideal structure for a n amorphous material
has been called a continuous r a n d o m network in which a complete amorphous
network of tetrahedrally bonded atoms can be formed, retaining the short-
range order except for small deviations in bond angles. The existence and
properties of hydrogen complexes in a-Si:H are discussed by Jackson and Zhang
(1990).
Native Defects
In this context a native defect in amorphous silicon can be seen as a departure
from the tetrahedral coordination of some particular Si atom t o four other Si
atoms: a broken Si bond, or a dangling bond. These defects are common to all
samples of amorphous silicon prepared around the world, and are identified by
a characteristic electron spin resonance (ESR) g = 2.0055 (Brodsky and Title,
1969; Taylor, 1984).
Actually ESR can detect only the paramagnetic species of dangling bond
with one unpaired electron; other possible states with zero or two unpaired
electrons are diamagnetic. The charge state of the defect in a particular mate-
rial depends on the location of the Fermi energy in that material; if the Fermi
energy is high in the gap then most defects will be in the diamagnetic D-
state occupied by two electrons, whereas if the Fermi energy lies close to the
valence edge, most defects will be in the diamagnetic D+ state, unoccupied
by electrons. When the Fermi energy lies close to the middle of the gap, as it
does in undoped, high quality a-Si:H, most defects will be in the paramagnetic
Do state with one unpaired electron. Both those defects that are present in
the dark after long times of thermal annealing, the so-called built-in defects,
and those that are induced by photoexcitation have the same ESR behavior.
74 Photovoltaic Materids
Another issue of considerable importance (see Sec. 3.4) is whether the de-
fects are intrinsic to the nature of the amorphous material, or whether they
are directly related to some extrinsic property such as a physical or chemical
abnormality (Stafford and Sabisky, 1987; Redfield and Bube, 1990; Redfield,
1991; Stafford, 1991).
Metastable defects in doped a-Si:H have been investigated by photomod-
ulation spectroscopy (Chen et al., 1992; Kocka et al., 1993). In addition to
information on the dangling bond defect (with state 0: in the dark, becoming
0%- or D: after capturing a photoexcited electron or hole, respectively) in
undoped material as summarized above, deep defects have been investigated
associated with phosphorus doping in a-Si:H:P (with state 0; in the dark,
becoming Dg after trapping a photogenerated hole), and with boron doping
in a-Si:H:B (with state DL in the dark, becoming Dk after trapping a photo-
generated electron). Exposure of a-Si:H:P to light produces both D p defects
and Di defects.
Doping with Standard Dopants (Beyer et al., 1977; Austin et al., 1979)
Doping a-Si:H with standard Si dopants such as phosphorus (n-type) or boron
(ptype) leads to an increase in the density of dangling bond defects if the
dopants are uncompensated, but actually to a decrease in defect density if
they are compensated (Amer and Jackson, 1984).
0
10’
z
Q
/-
I, 0 Boron
I I
U-
/ 0 Phosphorus
A Compensaled-partial
A Compensated-full
($Typical Error Bar 1’
I I I
10’ 9
10’6 10’’ 1olB 1c
N (Defects/cm3)
Fig. 3.5. Correlation of the densities of light-induced defects A N with the density of defects
N present in the starting a-Si:H material for various doping conditions. (Reprinted with
permission from A. Skumanich et al., Phys. Rev. B 31, 2263 (1985). Copyright 1985,
American Institute of Physics, College Park, MD.)
76 Photovoltaic Materials
lo'*] I
Fig. 3.6. (a) Comparison of experimental d a t a points from Park et al. (1989) for defect
density ( ~ m - ~in) a-Si:H with curves calculated from a stretched-exponential description
of optical degradation with a single fitting constant for all four curves ( 0 ) Sample A44, HI
(high-intensity Krf laser light), 343 K; (0)sample A62, HI, 343 K; (A) sample A44, LI
(lower-intensity tungsten-halogen light) 308 K; (0)sample A62, LI, 308 K. (b) T h e d a t a of
Park et al. (1989) given in (a), plotted using the form of Eq. (3.4) t o test t h e closeness of the
data to stretched exponentials, which produce straight lines in this kind of plot. (Reprinted
with permission from R. H. Bube et al., Appl. Phys. Lett., 5 7 , 79 (1990). Copyright 1990,
American Institute of Physics.)
logf*(N) versus logt yields a straight line with slope p and intercept log K .
Such stretched-exponential descriptions appear to be appropriate for a variety
of defect kinetics data obtained in recent years (Bube and Redfield, 1989).
78 Photovoltaic Materials
CONFIGURATION COORDINATE
Fig. 3.7. Simplified configuration-coordinate diagram that is generally used for the dangling-
bond defect in undoped a-Si:H. The minimum on the left represents the ground state, and
that on the right represents the dangling-bond metastable state.
80 Photovoltaic Materials
Weak-Bond Breaking
Some variation of the concept of weak-bond breaking underlies most of the
models describing the origin of metastable dangling bond defects in a-Si:H. It
appears to be a natural inference from the fact that the normal Si-Si bond
strength is about 2.2 eV, but it is known that photoexcitation can produce
dangling bonds with photons with energy as low as 1 eV. The first, and perhaps
the simplest model, for the formation of dangling bonds is one in which weak
bonds are broken by the energy released by the recombination between free
electrons and holes formed by photoexcitation (Adler, 1984; Stutzmann et al.,
1985). This model predicted a variation of defect density as t1I3during optical
degradation, but later examination showed that the apparent agreement of
the data, more completely described by a stretched exponential as described
in Sec. 3.3, with such a t1/3 behavior was coincidental (Redfield, 1989; Bube
e t al., 1990; Bube and Redfield, 1992).
Standard DDS
Defect Pool
. ~257meV
Fig. 3.8. Calculated densities of states in the gap of a-Si:H using the defect-pool model
approach, showing the difference between the simple two-peaks (above), similar to that
obtained for a Gaussian distribution of energies for a single type of defect, obtained when
the correlation energy U is large, and the “defect-pool” result (below) when U is smaller.
(Reprinted with permission from G. Schumm, Phys. Rev. B 49, 2427 (1994). Copyright
1994, American Institute of Physics, College Park, MD.)
where (Si-Si),b is the initial, weak bond state; SikH is a separate complex of
a Si atom that is bonded to one H atom; and Do is a neutral dangling bond
state formed when the weak bond breaks.
A comparison is given in Fig. 3.8 between two energy distributions of the
density of dangling bonds calculated according to the same defect-pool formu-
lae, on top when u << U , and at the bottom when IJ E U , where IJ is the
width of a Gaussian distribution of energies, and U is the defect correlation
energy (the energy difference between when the defect is negatively charged
and when neutral, assumed the same for all defects t o be of the order of 0.2 to
0.3 eV). The results shown on top in Fig. 3.8 are quite similar t o those expected
82 Photovoltaic Materials
Impurity-Induced Defects
It is known that uncompensated doping of a-Si:H increases the density of
metastable defects, as mentioned earlier. One of the early models for meta-
stable defects, proposed by Street (1982b), involved the formation of defects by
the process of self-compensation analogous to the behavior in partially ionic,
crystalline materials, in which, for example, acceptor-type cation vacancies
might be formed to lower the total energy when donor impurities are incorpo-
rated. Experimental tests of this model of doping have been reported by Kroetz
et al. (1991). It is uncertain, however, how similar such processes in a-Si:H
should be considered as compared to those in crystalline materials, and how
important such self-compensation processes are even in crystalline materials
in view of developments in recent years (Neumark, 1992; Chadi, 1994).
a-Si:H. Figure 3.9 illustrates the RTS model; in this case the ground state of a
foreign atom a t a substitutional site has fourfold coordination, and the center
is electronically inert; in the metastable state one bond to a neighboring Si
atom is broken, and the bonds of the foreign atom rehybridize to form three-
fold coordination. A critique of the RTS model includes the fact that the
DX-type center that was used as its basis does not occur in crystalline Si,
and particularly that dangling bond defects in a-Si:H occur in good-quality
undoped material in which the density of impurities (0: 2 x l O I 5 ~ m - C: ~ ,
7-10 x 1015 cmP3, and N: 5 x 1014 ~ m - is ~ much
) less than the density (5 x
of )dangling bond defects (Nakata et al., 1993; Kamei et al., 1996).
1017 ~ r n - ~
Fig. 3.9. A proposed RTS model for the metastable center based on a foreign atom in
analogy with the DX center in crystalline compounds; (a) ground state, (b) metastable
state. (Reprinted with permission from D. Redfield and R. H. Bube, Phys. Rev. Lett. 6 5 ,
464 (1990). Copyright 1990, American Institute of Physics, College Park, MD.)
Carrier-Induced Model
Some of the characteristics of dangling-bond defect formation in a-Si:H do not
appear to fit with the general theme of the above models, i.e. that recombi-
nation energy is the driving force. Typical examples are (a) the formation of
defects in accumulation layers where carrier recombination is negligible, (b)
photoinducing of defects as effectively at temperatures as low as 4 K as at
300 K , and (c) enhancing the annealing rate of excess defects by the addition
of photoexcitation (Street, 1991; Meaudre and Meaudre, 1991a,b; Isomura
and Wagner, 1992; Gleskova et al., 1993a,b). An open-ended proposal has
been advanced (Redfield and Bube, 1996) in terms of the configuration coor-
dinate diagram of Fig. 3.10, which resembles that for the EL2 defect in GaAs
(Dabrowski and Scheffler, 1989).
84 Photovoltuic Materials
QO Q1
(1
00
CONFIGURATION
a, 01 *
Fig. 3.10. (Lower) A proposed configuration-coordinate diagram for a carrier-induced defect-
formation mechanism . It is assumed that the conduction band is higher at a defect than
elsewhere, and that at E3 the latent center on the left can convert t o the defect on the
right with no energy barrier, simply by capture of a carrier from the conduction band.
The photoinduced process starts with t h e excitation of an electron at an existing defect;
the excited electron can move to a latent center t o form another defect. (Upper) Atomic
arrangements at configurations Qo, the latent-defect configuration, and & I , the dangling-
bond configuration. (Reprinted with permission from D. Redfield and R. H. Bube, Phil.
Mug. B 74, 309 (1996). Copyright 1996, Taylor & Francis Ltd., London.)
Fig. 3.11. Simple discrete-energy-level model for a trivalent defect, such as a dangling bond
defect in a-Si:H, with possible charge states of -, 0, and+. The dashed portion of each defect
level indicates t h e condition in which the level is in its state of lower electronic occupancy.
The arrows represent possible optical or thermal excitation energies, and recombination
energies as described in the text.
1 2 3 4 5 6 7
DEFECT DENSITY
Fig. 3.12. The Ferrni energy as a function of the defect density (in units of 10l6 and
measured at room temperature) at three different measurement temperatures: 298, 338 and
368 K , for a sample of undoped high-quality a-Si:H. At each of these temperatures data
are plotted from optical degradation versus time runs at 45OC (o), 8OoC (0)95OC ( o ) ,
and llO°C (A). The lines are empirical least-square linear fits to all the data at a fixed
temperature. (Reprinted with permission from R. H. Bube et al., J . A p p l . Phys. 75, 1571
(1994). Copyright 1994, American Institute of Physics.)
Amorphous Silicon 87
1 2 3 4 5 6 7
TOTAL DEFECT DENSITY
Fig. 3.13. Calculated variation of the densities (all densities in units of 1OI6 ~ r n - ~of) two
kinds of trivalent defects as the defect densities are increased with optical degradation, as
required to fit the d a t a of Fig. 3.8 using the model of Fig. 3.7 for two kinds of defects. Open
points are calculated for the higher-lying defects, and solid points are calculated for the
lower-lying defects. Measurements of Fermi energy were made at 298 K (o),338 K (O), and
368 K (A). (Reprinted with permission from R. H. Bube et al., J . A p p l . Phys. 75, 1571
(1994). Copyright 1994, American Institute of Physics.)
(3) Using the two-defect model, the variation with optical degradation of defect
densities, photoconductivity, and Fermi energy was used to obtain informa-
tion about the values of the electron capture cross sections of charged and
neutral dangling bond defects at room temperature. The following values
were obtained: Sy = (1kO.09) x cm2, S: M (0.6k0.05) x cm2,
Sf M (1.8 f 0.5) x cm’, and S: M (18 i 6) x cm2.
These values for capture cross sections are reasonably consistent with
earlier attempts to determine them. Measurements of the neutral cross
Amorphous Silicon 89
10'
7-
'0 4 0
10'
10- 5:-
10- ?--
10- 't
0.5 0.7 0.9 1.1 1.3 1 i
(4
Fig. 3.14. (a) Variation of the photoconductivity in a-Si:H as a function of the loca-
tion of the Fermi energy as varied by doping. The solid circles are experimental data
from LeComber and Spear (1986), and the open circles are the results of applying the
two-defect model of Fig. 3.13, using a typical variation of total defect density with
doping from Street (1985). (b) Comparison of the measured values of electron lifetime
( x lo-' s) corresponding to optical degradation at 353 K (solid circles ) with the values
of electron lifetime calculated from the two-defect photoconductivity model of Fig. 3.13
using measured total defect densities and the capture cross sections determined as above
(open circles). (Reprinted with permission from R. H. Bube e t al., J . Appl. Phys. 79,
1926 (1996). Copyright 1996, American Institute of Physics.)
90 Photovoltaic Materials
0 1 2 3 4 5 6
log t, sec
(b)
(4) Using the two-defect model and the capture cross sections just described,
comparison with different kinds of experimental data showed excellent cor-
relation. Figure 3.14(a) shows the dependence of photoconductivity on
Fermi energy according to the data of LeComber and Spear (1986), which
fits with the two-defect model using typical total defect densities as a func-
tion of doping level from Street (1985), considerably better than for the
single trivalent model mentioned above (Bube and Redfield, 1989). And
Fig. 3.14(b) shows a close fit between the experimental optical degrada-
tion of lifetime at 353 K and the values of electron lifetime calculated from
the photoconductivity model with the two trivalent defects and inferred
capture cross sections.
Amorphous Silacon 91
Fig. 3.15. An energy band diagram for a pi- cell in the short-circuit mode. (Reprinted
from D. E. Carlson, “Solar Cells” in Semiconductors and Semimetals. Vol. 21. Hydrogenated
Amorphous Silicon, Part D. Device Applications, 3 . I. Pankove, ed., p. 7 (1984). Copyright
1984, Academic Press, Orlando, FL.) ’
92 Photovoltaic Materials
El IT0 ( - 7 0 n m )
- - - - - ---
n-LAYER ( - 8 n m )
UNDOPED a-Si’H(-0.5pm)
p-LAYER (-20nm)
_-__--
STEEL
(b)
----
Fig. 3.16. (a) A pi-n solar cell structure on a glass substrate. (b) A pi+ solar-cell
structure on a steel substrate. (c) A stacked or multijunction structure where the i* layer is
a narrow-band-gap alloy such as a-Si:Ge:H. (Reprinted from D. E. Carlson, “Solar Cells” in
Semiconductors and Semimetals. Vol. 21. Hydrogenated Amorphous Silicon, Part D. Device
Applications, J. I. Pankove, ed., p. 7 (1984). Copyright 1984, Academic Press, Orlando, FL.)
Amorphous Silicon 93
- hv
GLASS
...
1 I l l 1
t t t t t
' i np i' n
(c)
Fig. 3.16. (Continued )
Fig. 3.17. (a) Optical absorption as a function of photon energy for undoped a-Si:H (solid
curve), p t y p e a-Si:H (dashed curve), and p t y p e a-Si:C:H (dashed-dotted curve.) Both
p t y p e films contain a few atomic percent of boron, and the a-Si:C:H film also contains
20 at. % of carbon. (b) Optical absorption coefficient as a function of photon energy for
a-Si:H, a-Sio.ls:Geo.sz:H (von Roedern et al. 1982), and a-Sio.7~:Co.zs:H(Morimoto et al.,
1982). (Reprinted from D. E. Carlson, "Solar Cells" in Semiconductors and Semimetals.
Vol. 21. Hydrogenated Amorphozls Silicon, Part D. Device Applications, 3 . I. Pankove, ed.,
p. 7 (1984). Copyright 1984, Academic Press, Orlando, FL.)
Amorphous Silicon 95
(a) D i n structure
..
(b)p-SiCh n a-SiN a-SiSn
a-SiC a-SIGe
n
Fig. 3.18. Schematic structure of a variety of cells involving a-Si:H. (Reprinted with permis-
sion from E. Maruyama et al., Solid State Phenomena 44-46, 863 (1995). Copyright 1995,
Scitec Pub., Switzerland.)
condition. High efficiency requires a diffusion length several times larger than
the thickness of the active region, which in turn requires a mobility lifetime
product p r > cm2V-' for a thickness of 0.5 pm, a condition that is met
in good-quality a-Si:H.
Large diffusion lengths in a-Si:H require a low density of defect states in the
gap. Such defects are most commonly associated with dangling-bond defects
as described in Sec. 3.4. They may arise during film growth by hydrogen out-
diffusion (Fritzsche et aZ., 1978), or may be associated with microstructural
imperfections such as polymer chains (Knights et al., 1979), or with impurities
such as oxygen (Pontuschka et aZ., 1982) and carbon (Morimoto et al., 1982). It
has been suggested that transport in a-Si:H may be strongly affected by inho-
mogeneities that give rise to potential fluctuations (Overhof and Beyer, 1981).
Metastable dangling-bond defects formed by photoexcitation (see Sec. 3.4) also
decrease the diffusion length.
High doping levels in the n- and p-type layers of a pi- a-Si:H solar cell
are desired to order to obtain a large built-in potential, approaching the band
gap in magnitude. Such high doping levels, however, are limited both by the
existence of the band-tail states and by low doping efficiency in a-Si:H, pre-
sumably because most dopant atoms do not occupy electronically active sites.
Estimates of the doping efficiency are in the range of 0.1 to 1.0% (Faughnan
and Hanak, 1983; Dresner, 1983). The Fermi level is about 0.2 eV below the
conduction band in a-Si:H:P (Spear, 1977) and about 0.5 eV above the valence
band in a-Si:H:B (Jan et al., 1980), limiting the built-in potential of a-Si:H
pi-n solar cells to about 1.0 eV (Williams et al., 1979). This built-in po-
tential is increased by using p-type a-Si:C:H layers (Tawada et d., 1982), but
the resistivity of the layer increases with the band gap, i.e. with the C/Si ra-
tio. The conductivity of the p-type layer in a boron-doped a-Si:H layer can be
significantly increased by forming microcrystalline-doped Si:H films (Matsuda
et al., 1980) but without significant increase in the built-in potential (Carlson
and Smith, 1982). The n-type layer can also be improved by making it micro-
crystalline (see Sec. 2.8), which is formed if the silane gas is heavily diluted
with hydrogen (Smith, 1991). The microcrystalline material has a higher con-
ductivity than a-Si:H, thus reducing resistive loss and improving the fill factor.
Under the assumption that the electric field in the i-layer is almost uniform,
if both the trapped charge and the free-carrier space charge are assumed to be
negligible, a simple model was developed that related the fill factor of pi-n
cells to the collection length (Crandall, 1982; Faughnan and Crandall, 1984)
with the results shown in Fig. 3.19. They found that the collection length at
Amorphous Silicon 99
,-
--
I I L
01 1 .o 10 100
Lc /d
Fig. 3.19. Dependence of the fill factor as a function of collection length for a-Si:H p i - n
cells. Comparison of theoretical variation (dashed line) and experimental points on a number
of pz-n solar cells. (Reprinted with permission from B. W. Faughnan and R. S. Crandall,
Appl. Phys. Lett. 44, 537 (1984). Copyright 1984, American Institute of Physics.)
AM1 EXPOSURE ( h r )
Fig. 3.20. Efficiency as a function of exposure time to AM1 radiation for a-Si:H solar cells.
Curves A and B are data for cells of the type shown in Fig. 3.16(a), and curve C is for a cell
of the type shown in Fig. 3.16(b). For curves A and B, the p t y p e region was a-Si:C:H, with
the carbon content for curve B (1.7 x lo2* ~ m - ~ being
) about 100 times larger than that
associated with curve A. (Reprinted from D. E. Carlson, “Solar Cells” in Semiconductors
and Semimetals. Vol. 21. Hydrogenated Amorphous Silicon, Part D. Device Applications, J.
I. Pankove, ed., p. 7 (1984). Copyright 1984, Academic Press, Orlando, FL.)
Amorphous Silicon 101
a-Si:H cells at about 200°C for several minutes. It was reported early that the
decrease in a-Si:H solar cell efficiency with time of photoexcitation was similar
to the formation of metastable defects in thin film a-Si:H under photoexcitation
(Staebler et al., 1981). The formation of new defects by exposure to light
decreases the lifetime, diffusion length, and drift length. It was also found
that the net space charge often increases with photoexcitation (Carlson et al.,
1983), contributing by itself to a decrease in efficiency.
4.0.
-2 3.0.
>
u
2.0'
-
0
LL
& 1.0
0.0 '
I
0
- .-
200
.
400 600
I . LAYER THICKNESS (nrn 1
I I
10’ 102 Id 1o4
Time (sec)
Time (sec)
(b)
Fig. 3.22. Efficiency of a-Si:H solar cells as a function of time of exposure t o light. Points
are the measured efficiencies, normalized t o their initial values, and curves are fits t o the
data using stretched exponentials. In (a) a fixed intensity of 140 suns was used with various
temperatures, whereas in (b) and (c) fixed temperatures of 5OoC and 100°C, respectively
were used with various intensities. (Reprinted with permission from L. Chen and L. Yang,
J. Non-Cryst. Solids 137/138,1185 (1991). Copyright 1991, North-Holland, Amsterdam.)
104 Photovoltaic Materials
} T=lOo'C -P-
(c)
(Redfield and Bube, 1991), based on annealing kinetics, existence and proper-
ties of saturation, degradation kinetics, thickness dependence of degradation,
and performance modeling of films and cells, with the conclusion that degra-
dation in a-Si:H solar cells is an i-layer effect directly related t o the formation
of metastable defects.
A number of other investigations have concluded that the degradation of
solar cell performance is directly related to the increase in the density of
metastable defects under illumination. (1) The relationship between fill factor
and carrier collection length shown in Fig. 3.19 has been tested by relating the
collection length to the defect density, with the result that changes in solar cell
performance due to changes in fill factor are directly correlated with defect
density (Frammelsberger e t al., 1991). (2) A computer model of cell degrada-
tion (Pawlikiewicz et al., 1991) showed good correlation between measured fill
factor and bulk defect density. (3) A comprehensive computer model of simi-
lar Schottky barrier MIS cells, for different i-layer thicknesses and durations of
light exposure, showed good agreement between observed degradation and cal-
culated quantum efficiency spectra based on the density of defects only (Malone
e t al., 1991). (4) Measurements of efficiency versus time of optical degradation
for large monolithic solar-cell modules involving a-Si:H show behavior quite
similar to that for films and single cells (Luft et al., 1991). (5) Optical degra-
dation at 60°C and modeling of a-Si:H pz-n cells well into saturation of their
Amorphous Silicon 105
determined and limited by the density of charged dangling bonds that are not
measured by ESR (von Roedern, 1993). (7) Significant differences in the stabi-
lized performance are reported depending on the conditions used to illuminate
a-Si:H multijunction modules (von Roedern et al., 1995). (8) Empirically a
relationship has been found between a greater electric field dependence of the
drift mobility (pd increases with electric field) of a n a-Si:H film in the annealed
state and a poorer stability of the photoconductivity upon photoexcitation
(Tang et al., 1996). (9) Along with the decrease of photoconductivity and
lifetime during exposure t o light, a continuous decay of the drift mobility was
also found, suggesting enhanced scattering associated with charged optically
induced defects (Tang and Braunstein, 1996). It is reported that different gen-
eration kinetics are measured using stretched-exponential kinetics for defects
leading to increased recombination (decrease of lifetime) and those leading to
decrease of drift mobility.
Shibata et al. (1988) used the technique of “&-doping”to increase the peak
boron concentration to 2 x loz1 cmP3 in the p t y p e layer, with apparently
beneficial results, producing a cell with initial efficiency of 11.5% The &doped
player consists of one or several very thin boron layers (0.1 to 0.5 atomic layer)
and undoped a-Si:H layers.
Measurements of electron transport and recombination (Nebel et al., 1988a;
Bauer et al., 1988), and diffusion lengths from optical grating techniques (Nebel
et al., 1988b) on a-Sil-,:Ge,:H alloys have been reported for values of z be-
tween 0 and 0.3. It is reported that the electron mobility in a-Si:Ge:H decreases
with increasing hydrogen concentration and germanium concentration (Fort-
mann et al., 1991). Capacitance techniques have been applied to investigate
the midgap density of states in a-Si:H and a-Si:Ge:H p i - n solar cells and
Schottky junctions (Hegedus and Fagen, 1992).
Optical degradation has been observed for a-Si:Ge:H films (Aljishi et al.,
1987) with effects depending on the band gap of the alloy. If the band gap is
larger than 1.4 eV, an increase in subgap absorption, a decrease in photocon-
ductivity, and a decrease in dark conductivity with an increase in activation
energy was observed. For band gaps less than 1.4 eV, an increase in subgap
absorption, constant photoconductivity, and an increase in dark conductivity
with a decrease in activation energy was observed. Chu et al. (1988) carried
out a similar investigation concentrating on band gaps between 1.2 and 1.4 eV
of interest for multijunction cells.
Variations in deposition techniques have also been tested with a-Si:Ge:H
alloys. Weller et al. (1987) used dc-glow discharge, and Hegedus et al. (1988)
used photo-assisted CVD.
Of particular interest is the series stack of two or three solar cell junctions
with alloys a-Si:C:H with larger band gap, a-Si:H itself, and a-Si:Ge:H with
smaller band gap, shown in Fig. 3.18(f) and in more detail in Fig. 3.23. Since
the i-layer thickness in these cells is narrower than in a simple a-Si:H cell,
it is expected that optical degradation effects would be reduced. Offsetting
this gain is the apparent increased instability of a-Si:C:H compared to a-Si:H,
and uncertainty about the apparent decreased instability of a-Si:Ge:H alloys
because of cited evidence that there is a lack of correlation between defect
density in the bulk and stability of these devices (Xu et al., 1993). Still,
research on the effects of H, Si, and Ge in alloy films has been used to produce
what is claimed to be the most stable a:Si:Ge:H solar cell (Terakawa et al.,
1994). A variety of approaches are being considered to achieve more stable
multijunction cells (Haku, 1991; Maruyama et al., 1993a,b; Hishikawa et al.,
1994; Sayama, 1994).
108 Photovoltaic Materials
5 light
a-Si -/
Types of heterojunction solar cells involving both a-Si:H and other forms of Si
promise increased efficiency.
1 1 1 1 1
. \ ,-
TCopper
7 Indium
Diselenide
Glass
Fig. 3.24. Solar cell structure for a 14.2% efficient multijunction cell consisting of an a-Si:H
pin cell and a ZnO/CdS/CuInSe2 cell. (Reprinted with permission from D. L. Morel,
Solar Cells 24, 157 (1988). Copyright 1988, Elsevier Sequoia S. A., Lausanne, Switzerland.)
CHAPTER 4
4.1. Overview
It is useful to place gallium arsenide and other 111-V materials relevant to solar
cell applications in the context of the Periodic Table of the Elements and their
relationship to the element silicon that formed the basis for the previous two
chapters. A pertinent subsection of the Periodic Table is shown in Fig. 4.1.
Noting the usual trend that elements and compounds have increasing ionic
binding and band gaps as the rows lie higher in the periodic table, and as the
difference between the columns increases, we see the effect described in the
previous chapter that the band gap of Si, in the same column as Ge but lying
5B 6C 7N 8 0
1 3 A Y i y P 1 6 s
29 Cu 30 Zn 31 Ga 32 Ge 33 As 34 Se
47 Ag 48 Cd
/In 49 50 Sn 51 Sb
\ Te 52
79 Au 80 Hg 81 T1 82 Pb 83 Bi 84 Po
Fig. 4.1. A subsection of the Periodic Table of the elements showing the relationship between
Group IV elements like Si and Ge, Group 111-V compounds like GaAs and InP, and Group
IILVI compounds like CdTe.
111
112 Photovoltaic Materials
one row higher, is greater than the band gap of Ge. GaAs consists of Ga in
the column to the left of Ge, and As in the column to the right of Ge, giving
rise to a band gap in GaAs, from these considerations only, which is greater
than that of Ge, and would be expected to be more like the band gap of Si.
Actually, since GaAs is a direct-band gap material, its band gap (1.43 eV)
is somewhat larger than that of crystalline Si (1.1eV) with an indirect band
gap. It may be concluded, therefore, that GaAs and InP (which makes up for
In lying in a lower row than Ga, by P lying in a higher row than As) are the
111-V compounds most like Si. We note in passing that if we extend our choice
diagonally down to the left from Si through Ga, we reach Cd, and if we extend
our choice diagonally down to the right from Si through As, we reach Te; we
recognize the fact that CdTe is the major 11-VI material of interest for solar
cells, as discussed in Chapter 5 .
t
ci
d
Fig. 4.2. Summary of band gaps and lattice constants for a number of zincblende 111-V com-
pounds, compared with diamond structure Si and Ge, and zincblende ZnTe, CdTe, and HgTe.
Gallium Arsenide and Other III-V Materials 113
The direct band gap of GaAs is near the optimum for solar conversion, as
shown in Fig. 1.2, and leads to 97% absorption of AM1 radiation in a thickness
of about 2 pm. Its higher carrier mobilities than Si allow the fabrication
of high-frequency devices, and it forms a variety of lattice-matched ternary
compounds allowing for controlled variations of properties. GaAs solar cells
should be operable at higher temperatures than silicon cells, and since they
may be very thin, GaAs solar cells are expected t o be radiation resistant. The
band gap versus lattice constant relationship is given in Fig. 4.2 for eight 111-V
compounds with zincblende structure, as well as for a few other semiconductors
for reference such.as Si and Ge with the diamond structure, and ZnTe, CdTe
and HgTe with zincblende structure. For example, Fig. 4.2 shows the close
lattice match between AlAs, GaAs and Ge for three materials with a wide
range of band gaps, as well as for AlSb, GaSb and InAs.
c
z
W
0
a
W
n
>.’
0
z
w
0
LL
LL
W
Fig. 4.4. Representative values of GaAs-based solar cell efficiency as a function of year.
Three types of cells are included: ( 0 ) homojunction, (m) heteroface (buried homojunction)
without concentration, and (A) heteroface with concentration. Details are discussed further
in Sec. 4.3.
Efficiencies of GaAs-based solar cells are among the highest of any solar
cells, lying between 20% and 30%. Representative values of efficiency as a
function of year and type of cell are plotted in Fig. 4.4 and are discussed
further in Sec. 4.3. Because of the relatively high cost of the material and of
cell fabrication, the development of GaAs-based (and InP-based) cells has been
strongly directed toward concentrator use and space applications. General
issues related to GaAs-based solar cells are discussed in several publications
(Fahrenbruch and Bube, 1983; Fraas, 1985; Bube, 1990, 1993; Boes and Luque,
1993).
Gallium Arsenide and Other III-V Materials 115
-
-
-
-
-
SERIES RESlSTAmE VHUES Wovn IN n cm2
0.w I ~ . . . I . . . . I . . .._
1.00 1o.m loo.w lm.w 1.W4
CONCENlPATIW
Fig. 4.5. Calculated solar efficiency as a function of concentration ratio for an AlGaAs/GaAs
LPE cell for various values of lumped series resistance. (Reprinted with permission from
R. Sahai et al., 13th IEEE Photovoltaic Specialists Conference, p. 946, (1978). Copyright
1978 IEEE, NY.)
FRONT CONTACT
AR COATING
0.6v CAP
3 . 2 ~n GaAs BASE
O l v n AIozGao& BACKWINWW
0.51” n+ GaAs
SUBSTRATE
t BACK CONTACT I
Fig. 4.6. Solar cell structure for a 24.0% efficient GaAs solar cell. At 1 sun, & = 1.05 V,
Jsc = 27.1 mA/cm2, fill factor = 0.845. (Reprinted with permission from K. A. Bertness
et al., 20th IEEE Photovoltaic Specialists Conference, p. 769 (1988). Copyright 1988 IEEE,
NY .)
0
3.
I
0.1
0.502
I
0.2
9.8
I
0.3
I
0.4
VOLTAGE, V
;;!0.5 0.6
Fig. 4.7. Light current/voltage curves for Au/GaAs Schottky barriers for various treatments
of the GaAs surface before application of the metal barrier contact. No antireflection coating
is present. (1) “Clean” interface, (2) GaAs surface exposed to air at 300 K for 4 hr, (3)
exposed t o air at 300 K for 95 hr, and (4)exposed t o air at 403 K for 70 hr. (Reprinted with
permission from R. J. Stirn and Y. C. M. Yeh, Appl. Phys. Lett. 2 7 , 95 (1975). Copyright
1975 American Institute of Physics.)
Gallium Arsenide and Other III-V Materials 121
through the use of a focused laser beam as the excitation source. Using a
theoretical model to interpret measured photoluminescence decay transients,
minority carrier properties in GaAs solar cells can be determined at different
stages of processing. The measurements also provide minority carrier param-
eters, such as lifetime and interface recombination velocity, useful for device
modeling.
MIS Cells
The performance of Schottky barriers on GaAs can be considerably enhanced
by the inclusion of an interfacial insulator layer between the metal and the
GaAs to form an MIS device. Figure 4.7 shows the striking dependence of
open-circuit voltage in Au/oxide/GaAs MIS junctions on the introduction of
an insulating oxide layer (Stirn and Yeh, 1975, 1977; Stirn et al., 1977). High-
efficiency cells can be made either by water-vapor thermal oxide growth or by
deposition of Sb203 onto suitably etched GaAs surfaces. These were called
AMOS cells (antireflection-coated metal oxide semiconductor) and gave effi-
ciencies up to 17%.
It is the increase in diode factor A caused by the oxide layer that is the
primary contributor t o the increased open-circuit voltage. For AMOS cells on
(100) GaAs surfaces, oxidation causes a n increase in A from 1.0 to 1.2 without
an increase in J,. On (111) GaAs surfaces, on the other hand, J , increases on
oxidation by about an order of magnitude, and diode factors of 1.4 to 1.6 are
found.
An n-i-p structure for a GaInP solar cell with an efficiency of 15.7% un-
der AMO, among the highest efficiencies of GaInP single junction cells, was
achieved by introducing an undoped, intrinsic GaInP layer a t the p n junction
to reduce the interdiffusion of Zn into the emitter, and separating the p and
n-regions (Tsai et al., 1995).
C L E F T Films
GaAs is the only material of current interest for solar cells that can be fabri-
cated in the form of a single-crystal thin film solar cell by a desirable process
(Gale et al., 1988, 1989). This possibility is made economically attractive by
the fact that the film can be grown on a reusable, single-crystal GaAs substrate
by the technique known as “cleavage of lateral epitaxial films for transfer”
(CLEFT). The CLEFT process does away with the necessity of using thick,
non-reusable, expensive GaAs substrates.
122 Photovoltaic Materials
The various steps in CLEFT fabrication are as follows (Fraas, 1985): (a)
the single-crystal substrate is coated with carbon before epitaxial layer growth,
(b) narrow slits are left in the carbon where the single crystal is exposed, (c)
epitaxy is nucleated in the slits and the similarly oriented crystallites grow
laterally rapidly across the carbon until they join to form a continuous film,
(d) the thin-film solar-cell layers are then grown, (e) contacts are added and
the film is glued to a glass superstrate, (f) the master substrate is removed by
cleaving and can be reused.
GaAs Alloys
In,Gal-,As concentrator cells have been prepared using MOCVD on GaAs
substrates, with a band gap of 1.15 eV for x = 0.25, and 1.35 eV for x = 0.07
(Werthen et al., 1988). With a possible long range application to multijunc-
tion cells, these single-junction cells showed efficiencies greater than 24% at
400 suns. The cell with x = 0.25 has about the same band gap as Si, but has
a direct band gap, and yields & = 0.80 V compared with 0.67 V for Si. Im-
provements in current collection should lead to considerably higher efficiencies
for the InGaAs cells.
1
0 llpm
4
ni AI.Gq .h
C M R
1
WINDOW
- 1
--___----
0 lSlrm n A10.376.0- ----- EMITTER
UPPER CELL
3Wrm P *lo.nwo- TTOM M E T A L CELL EASE
TOP METAL
OlJum p' AIxGalxh STOP ETCH
O U m P *10.37Gq(LJA* UPPER CELL BOTTOM CONTACT
N N
N
SUBSTRATE-
I
t
350 "rn p Gar
BACKCONTACT
I1
Fig. 4.8. Solar cell structure for a 23.9% efficient monolithically grown, two-junction solar
cell with an A I o . ~ ~ G ~ top. ~ cell s a GaAs bottom cell. At 1 sun, q50c = 2.41 V,
~ A and
Jsc = 14.9 mA/cm2 and fill factor = 0.84. (Reprinted with permission from G. F . Virshup
et al., 20th IEEE Photovoltaic Speczalists Conference, p. 441 (1988). Copyright 1988, IEEE,
NY.)
124 Photovoltaic Materials
r 1
n'GaAs
0.5 p n
- 10'8 ern.'
* I x 10)' 6.0 pm
I we Substrate -300p
Mrlnl
Si doping spike
n= 6 x 10’’ I
0.025
0.1
AllnP
GalnP
[si 1 n= 4 x 1017
n= 2 x 10’8 I (window)
I I I
Fig. 4.10. Solar cell structure for AM0 standard solar spectrum for a GaInP/GaAs monc-
lithic multijunction. (Reprinted with permission from K. A. Bertness et al., 24th IEEE
Photovoltaic Specialists Conference, p. 1671 (1994). Copyright 1994, IEEE, NY.)
the G a o . ~ 4 I n o . ~ 6 A s o . ~ ~can
P 0 . 3be
~ grown directly without the buffer layer
normally needed t o avoid the propagation of defects from the substrate. A
schematic of this proposed cell is given in Fig. 4.9. For space applications
Gao.6sIno.32ASo.34Po.66(band gap = 1.7 e v ) has demonstrated good radiation
resistance, with &c = 1.161 V, J,, = 28.9 mA/cm2, fill factor = 0.86 and
efficiency = 21.4% for AM0 illumination.
Record-breaking efficiencies have been reported for GaInP/GaAs two-
terminal multijunction cells: 29.5% for one-sun AM1.5, 30.2% for 160-suns
AM1.5, 25.7% for one-sun AMO, and 19.6% at AM0 after cm-’ 1-MeV
electron irradiation (Bertness et al., 1994). Efficiency advances were attained
126 Photovoltaic Materials
concentration (suns)
InP, like GaAs, is a direct band-gap material with a band gap of 1.34 eV, close
to the maximum for solar energy conversion. Also like GaAs, homojunction
cells are limited by surface recombination at the incident surface (Galavanov
et al., 1967), but heterojunctions with good lattice matching, and heteroface
junctions (buried homojunctions), are much more efficient. InP crystals are
grown by the Czochralski method at high pressures or by using a liquid encap-
sulation technique to preserve stoichiometry.
Heterojunction Cells
Highly efficient CdS/InP heterojunction cells have been fabricated by vac-
uum evaporation, chemical vapor deposition, and close-spaced vapor transport
(CSVT). For more details see Fahrenbruch and Bube (1983).
Gallium Arsenade and Other III-V Materials 127
AEc = 0.112 eV
6P = 0.088 eV
InP
2.42 eV = 6.1 10” cn-3
v 4
n - CdS
2.8 x 1019 cme3
Fig. 4.12. Energy band diagram for a CdS/InP heterojunction prepared by close-spaced
vapor transport of a CdS film onto single crystal InP. (Reprinted from A. Yoshikawa and
Y. Sakai, Solid State Electronics 20, 133 (1977). Copyright 1977, with kind permission from
Elsevier Science Ltd., The Boulevard, Langford Lane, Kidlington OX5 lGB, UK.)
Heteroface Cells
A heat treatment a t about 600°C in a non-oxidizing atmosphere improves the
CdS/InP cell described above, giving an efficiency of 14% (Shay et al., 1976),
and suggesting the formation of a heteroface junction, n-CdS/n-InPlpInP.
The efficiency of a CdS/InP cell was further increased t o 15% by using
chemical vapor deposition of CdS on InP using an open-tube H2S/H2 flow
system (Shay et al., 1977; Bettini et al., 1977, 1978). Once again the properties
of the cell suggest that it is an InP buried homojunction. Solar cell parameters
for AM2 illumination were = 0.79 V, Jsc = 18.7 mA/cm2, f f = 0.735.
In view of the importance of lattice matching for high efficiency that un-
derlay the above discussion of CdS/InP solar cells, it is perhaps surprising to
128 Photovoltaic Materials
InP
ITO
20 In‘ 16 17
nf- 1x10 n-1x10 p-3x10
find that cells with efficiency greater than 14% can be made from junctions
between indium-tin-oxide (ITO) and p t y p e InP for which the lattice mismatch
is large (Sree Harsha et al., 1977). Subsequent investigation (Bachmann et al.,
1979; Tsai et al., 1980) has shown that the high efficiency of the ITO/InP
junctions is the result of the formation of a buried homojunction in the InP
by the action of sputter deposition, with the resulting band diagram shown in
Fig. 4.13.
Homojunction Cells
A large scale commercial process has been described for the fabrication of n + p
homojunction InP solar cells for space power generation, involving the diffusion
of sulfur into p t y p e InP substrates in a sealed quartz ampoule (Okazaki et al.,
1988). The best 2 x 2 cm2 cell had q5,, = 0.828 V, J,, = 33.7 mA/cm2, fill
factor = 0.816 and efficiency = 16.6% for AMO.
In spite of the front surface recombination problems, and in the absence
of a suitable lattice-matched, wide-bandgap material that could be used as a
window layer for passivation of the front surface, a high efficiency n f p junction
has been prepared by MOCVD with a graded-junction, or front-surface-field
structure, to decrease front surface recombination losses (Keavny et al., 1990).
The cell consists of a p t y p e heavily-doped buffer layer, a lightly-doped p
type base layer, an n-type emitter, and an n-type Ino.53Gao.47As cap layer.
The n-type emitter is made up of two layers: A moderately heavily doped
region (3 x 10'' ~ m - near ~ ) the junction, and a very heavily doped region
(3 x lo" cmP3) at the surface. After metallization with Cr-Au-Ag, the cells
were mesa etched, and the InGaAs layer was removed from the active area
with a selective etch. Finally the cells had a ZnS/MgF2 two-layer antireflection
coating. A 4 cm2 cell for AM0 had & = 0.876 V, J,, = 36.34 mA/cm2, fill
factor = 0.824 and efficiency = 19.1%. The cells showed a degradation of 4.7%
after exposure to 1014 cmP2 1 MeV electrons.
Hydrogenation of n+-p InP solar cells has been shown to increase the effi-
ciency from 14.8% to 17.5% a t AM0 because of a reduction in carrier concen-
tration in the near-surface layer due to the formation of an acceptor-hydrogen
complex (Min et al., 1993).
MIS Cells
MIS cells with efficiency of 14.5% have been fabricated on InP, similar to those
described above for GaAs (Kamimura et al., 1981).
-
E InP/CdS SOLAR C E L L S
G aA s / S i
A record-breaking 31% efficient GaAs/Si mechanically stacked multijunction
(MSMJ) concentrator solar cell has been developed (Gee and Virshup, 1988).
Special care was needed to insure that the GaAs cell had maximum trans-
parency for sub-band gap radiation. InGaAs is a possible substitute for the Si
cell.
GaAs/CuInSez
An MSMJ four-terminal device, consisting of a GaAs thin film top cell made by
the CLEFT process and a ZnCdS/CuInSez thin bottom cell, gave an efficiency
of 21.3% (19.5% from the GaAs cell, and 2.8% from the CuInSe2 cell) (Stanbery
et al., 1987; Kim et al., 1988). Good radiation resistance makes these cells
candidates for flat-plate space applications.
GaInPz/GaAs
Efficiencies of about 25% have been reported for monolithic GaInPz/GaAs
multijunction cells with a tunnel-junction interconnect (Olson et al., 1989;
Kurtz et al., 1990), and plans for a GaInPz/GaAs/Ge triple junction have
been described (Chiang et al., 1994). Comparable efficiencies are obtained for
p-on-n GaInPz/GaAs two-terminal, monolithic multijunction cells (Sharps
et al., 1994).
A single junction InGaP solar cell with an n-ppf structure and a BSF
achieved an efficiency of 18.48%, and InGaP/GaAs monolithic multijunctions
were prepared with an efficiency of 27.3% (Takamoto et al., 1994; Kurita et al.,
1995).
132 Photovoltaic Materials
GaAs/GaSb
An efficiency of 34% has been achieved in an MSMJ GaAs/GaSb cell (Fraas
et al., 1990, 1991). Cells based on pn-InGaSbAsln-GaSb lattice matched
heterostructures have also been fabricated by LPE with the long-wavelength
edge of photosensitivity of the cells at about 2.15-2.2 pm (Andreev et al.,
1996).
InP/GaInAs
An efficiency of 22.2% has been achieved with a monolithic multijunction con-
centrator cell based on InP/GaInAs, and research has been carried out on
other variations in this system (Wanlass et al., 1991, 1994).
x 10-8
Fig. 4.15. Changes in damage constant K L for minority-carrier diffusion length for InP-
related materials and GaAs as a function of InP composition. (Reprinted with permission
from M. Yamaguchi et al., Japanese J. Appl. Phys. 34, 6222 (1995). Copyright 1995,
Japanese Journal of Applied Physics.)
5.1. Overview
When work on thin-film Cu,S/CdS cells came t o a n end because of the un-
avoidable degradation associated with the effects of Cu diffusion, the question
naturally arose as to what other p t y p e materials could form useful hetero-
junctions with CdS. Two such materials are InP, discussed in Chapter 4,and
CuInSe2 discussed in Chapter 6. It would also be desirable to find suitable
p t y p e materials among other 11-VI compounds, capable of being deposited in
thin film polycrystalline form in ways analogous to those used for CdS. Of the
six 11-VI chalcogenides (ZnS, ZnSe, ZnTe, CdS, CdSe, and CdTe) only ZnSe
and CdTe can be prepared in both n-type and p t y p e form, and of the rest only
ZnTe can be prepared in high-conductivity p t y p e form. For the preparation
of p-n heterojunctions from 11-VI chalcogenides, therefore, the choice of the
p t y p e member is essentially restricted to ZnTe or CdTe, since p t y p e ZnSe has
yet to be demonstrated in polycrystalline form. This leaves the nine potential
possibilities summarized in Table 5.1, together with their alloys (Bube, 1976).
The 11-VI compounds n-type CdS and n-type ZnO have been used for the
high-conductivity window-layer in several heterojunction cells and examples
are given in several places in this book.
The two most likely candidates for efficient solar cells are n-ZnCdSlpCdTe
and n-ZnSSelpCdTe. A plot of band gap versus lattice constant for the II-
VI chalcogenides and a few other materials of interest are shown in Fig. 5.1
(Bube and Mitchell, 1993). All of the 11-VI heterojunctions show appreciable
lattice mismatch except n-CdSelpZnTe, but such mismatch by itself does not
appear to be a serious problem for CdTe-based cells. The stoichiometry limits
of existence are shown in Fig. 5.1, although several of the compounds have
been prepared outside these limits in metastable form.
In all of the different ways in which solar cells have been fabricated based
on junctions involving p C d T e crystals and polycrystalline thin films, three
problems appear to be dominant: recombination losses associated with the
135
136 Photovoltaic Materials
E -
0 CulnSe,
CdTe
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
5
1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 199
Calendar year a m w
Fig. 5.2. Best one-of-a-kind laboratory cell efficiencies for polycrystalline thin-film cells
based on CdTe or CuInSez. (Reprinted with permission from K. Zweibel et al., 25th ZEEE
Photovoltaic Specialists Conference, p. 745 (1996). Copyright 1996, IEEE, NY.)
(7)CLEAVED
I I
Fig. 5.3. Flowchart summarizing the various CdTe surface treatments and their effects.
(Reprinted with permission from J.-P. Haering et al., J . V a c u u m Science and Technology
A l , 1469 (1983). Copyright 1983, American Institute of Physics.)
10-2 c
ld7
10.e
t, 01 02 03 04 05 06
FORWARD BIAS ( V )
(a)
I J(mAlcm2) f
(b)
Fig. 5.4. (a) Dark J-V characteristics of Cr/CdTe junctions formed on (110) surfaces:
cleaved (C), Br:MeOH-etched (E), and Br:MeOH-etched plus hydrogen heat treated (EH).
(b) Light J-V characteristics of Cr/CdTe junctions formed on (110) surfaces: cleaved (C),
Br:MeOH-etched (E), and Br:MeOH-etched plus hydrogen heat treated (EH). (Reprinted
with permission from J.-P. Haering et al., J. Vacuum Science and Technology A l , 1469
(1983). Copyright 1983, American Institute of Physics.)
C a d m i u m Telluride and Other II-VI Materials 143
Fig. 5.5. Light J-V characteristics of Cr/CdTe junctions formed on polished and Br:MeOH
etChed (PE), and polished, Br:MeOH etched and heat treated (PEH) surfaces of different
orientations. (Reprinted with permission from J. G. Werthen et al., J . Appl. Phys. 54,
5982 (1983). Copyright 1983 American Institute of Physics.)
Surface Oxidation
A high-resolution electron microscopy study of a CdTe single crystal, after
being stored in air at room temperature under room illumination for a period
of abut 30 days, showed the existence of a 60 nm thick TeOz layer on the (110)
surface (Ponce et al., 1981).
The effect of this surface oxide layer on the properties of Cr/CdTe junctions
formed on cleaved surfaces exposed to air for different lengths of time was
observed (Werthen et al., 1983a). An increase in open-circuit voltage observed
with length of time the cleaved surface was exposed to air was attributed to
the presence of the TeOz on the air-exposed surface.
Thermal oxidation df single-crystal p t y p e CdTe was carried out in dry and
wet oxygen at temperatures between 350°C and 500°C (Wang et al., 1987).
The growth rate was increased by about a factor of 2 by changing from dry
to wet oxygen, and was slightly higher on (111) Cd and (111) Te faces than
on (110) faces. The thickness of the oxide varied as the square-root of the oxi-
dation time, implying a diffusion controlled process with an activation energy
of 1.2 eV for thermal oxidation. Examination of the oxide by Auger electron
spectroscopy, x-ray photoelectron spectroscopy, and transmission electron mi-
croscopy showed that the composition of thick layers is CdTe03. A related
Cadmium Telluride and Other II-VI Materials 145
study of the solid-state quaternary phase equilibirum diagram for the Hg-Cd-
Te-0 system, including ternary diagrams for Cd-Te-0 and Hg-Te-0, indicated
that CdTeOs is the first oxide to form and remains stable (Rhiger and Kvaas,
1983).
Cr/CdTe and Cr/CdTe03/CdTe (MIS) junctions, involving the thermal
oxide CdTe03, were fabricated on both p and n-type CdTe single crystals
by oxidation of single crystal CdTe after polishing, Br:MeOH etching, and
hydrogen heat treatment (Wang et al., 1989). For n-type CdTe the oxide
increases the open-circuit voltage (to 0.71 V for a 1 nm thick oxide layer)
over that of the Schottky barrier (0.39 V), whereas for p t y p e CdTe the oxide
decreases the open-circuit voltage (from 0.47 to 0.42 V). Postulated energy-
band diagrams for the two types of junction are given in Fig. 5.6(a) for the
Cr CdTeO, p-CdTe
-
Evx
1.3
4.3
f
O.' f f
3 .O
3.2
<
t
-Ec
f 0.5
I
f
(4
Fig. 5.6. (a) Energy band diagram for a Cr MIS Junction on p C d T e crystal with native
CdTeO3 oxide as the insulator. (b) Energy band diagram for a Au MIS junction on n-CdTe
with native CdTe03 oxide as the insulator. (Reprinted with permission from F. F. Wang
et al., J . A p p l . Phys. 6 5 , 3552 (1989). Copyright 1989 American Institute of Physics.)
146 Photovoltaic Materials
Au cdTeO3 n-CdTe
(b)
Contacts
The work function of p t y p e CdTe is sufficiently large (of the order of 5.7 eV)
that no metal has a large enough work function to make an ideal ohmic con-
tact. In addition, the Fermi energy at the surface can be partially pinned by
surface states, depending on the surface treatment. The effort to obtain a
low-resistance contact with ohmic behavior has therefore focused on obtain-
ing a contact to highly-surface-doped p C d T e suitable for tunneling transport.
Low-resistance contacts to p C d T e have been reviewed by Ponpon (1985) and
Fahrenbruch (1987).
Cadmium Telluride and Other 11-VI Materials 147
Early work (Gu et al., 1975; Jager and Seipp, 1981) indicated that etching
the surface with K2Cr207:HzS04 provided the basis for a low-resistance con-
tact using Au or Ni-Au. Related results were obtained by Anthony et al. (1982)
who found that the best contact was a CuAu alloy, prepared by co-evaporating
Cu and Au from the same boat with Cu comprising approximately 12 at.%
of the evaporant, on a p t y p e CdTe surface etched with K2Cr207:HzS04 (a
CuAu/Chr contact). Auger profiling measurements indicated that there is a
marked enhancement of the Te/Cd ratio over a region of about 48 nm from
the surface. The presence of Cu in the contact increases the effective doping
density in the depletion region. The contact resistivity at room temperature
depends on the CdTe bulk resistivity, as shown in Fig. 5.7, but it increases
rapidly upon cooling, to a value greater than lo6 ohm-cm2 at 125 K. The
CuAu/Chr contacts were found to be essentially stable a t room temperature,
but increased in resistivity by a factor of about four when annealed between
100°C and 200°C. These contacts could be described by a thermally assisted
tunneling model (Padovani and Stratton, 1966).
Contact resistivities comparable to those of CuAu/Chr contacts can also
be obtained by the use of a material like p Z n T e as the contact material to
polycrystalline CdTe in the formation of p i - n junctions as described below
(Meyers, 1989; Nouhi et al., 1989), or ofpHgTe as the contact material because
of its larger work function than p C d T e (Chu et al., 1988). Contact resistances
of the order of 0.4-0.8 ohm-cm2 were obtained for deposition of HgTe, either by
direct combination of the elemental vapors in a gas flow system or by CSVT,
onto large-grain polycrystalline CdTe with resistivity of 200-400 ohm-cm. Thin
(< 100 nm) ZnTe:Cu films were deposited by an electrochemical method to
make contact to the polycrystalline CdTe in a CdTe/CdS/ITO solar cell, giving
higher optical transmission than similar vacuum evaporated ZnTe:Cu contacts
and leading to an 8.7% efficient cell (Mondal et al., 1991, 1992). Cells made
with electrodeposited CdTe and ZnTe:Cu vacuum-evaporated contacts t o the
CdTe have been reported with an efficiency of 12.9% (Tang et al., 1996).
Reproducible low-resistance contacts of the order of 0.01 ohm-cm2 at 30 K
and 0.15 ohm-cm2 at 77 K were achieved on Li-diffused single-crystal CdTe,
using n-butyllithium, with Au as the contact metal (Lee and Bube, 1983). Li
doping achieved a p+ surface layer with p = lo1’ cm-3 that made tunneling
currents possible to produce the low contact resistivities observed (from Au to
pf surface), independent of the bulk resistivity or even the bulk conductivity
type. Unfortunately the contact resistivity changes with time, first decreasing
and then later increasing, and with temperature, probably involving diffusion
of Li acceptors away from the surface and loss of Li acceptors by interaction
with lattice defects or by precipitation. Even so, contact resistivities less than
0.03 ohm-cm2 can be achieved at 300 K for periods of several months.
Subsequent research on these contacts showed that values of contact resis-
tivity as low as 2.1 x l o p 3 ohm cm2 at room temperature and 3.7 x lo-’ ohm
cm2 at liquid N2 temperature can be achieved on Li-diffused CdTe with Au
as the contact metal (Lee and Bube, 1985). An MIS tunnel diode model was
proposed to account for the observed current-voltage characteristics of the con-
tacts, in view of a surface analysis that indicated that the crystalline structure
of the CdTe surface had been greatly disturbed and a Te-rich surface layer
produced, and the degradation of the contacts with time is probably caused
both by diffusion of Li and by a decrease of interface state density through
out-diffusion of excess Te.
A contact process for making low-resistance contacts to polycrystalline
pCdTe has been claimed to be compatible with module fabrication needs
(McCandless et al., 1994). After CdC12 processing, the CdTe surface is coated
with a thin layer of Cu by electron beam evaporation that is then heated
and reacted, diffusing Cu into the CdTe to dope it p t y p e . The reaction re-
moves elemental Cu from the surface, that might lead t o chemically-induced
Cadmium Telluride and Other II-VI Materials 149
instability and produces a conductive surface that can be contacted using dif-
ferent materials. Using CdTe films from a variety of sources, qboc varied from
0.78 to 0.86 V, the fill factor from 0.71 to 0.77, and the efficiency from 11.0 to
14.1%. A related process involving the deposition of a very thin Cu,Te layer
on the CdTe surface has been described by Florez et al. (1996).
Although most polycrystalline-based devices are made with contacts that
involve Cu in some way, there are a number of alternatives. Ghosh et al. (1995)
reported making contacts with resistivities of about 0.1 ohm-cm2 by depositing
electroless Ni containing P onto a previously prepared surface.
C V D and M O C V D Doping
Chu et al. (1985) reported doping of p t y p e CdTe multicrystalline films de-
posited on non-crystalline substrates from the direct combination of the ele-
ments in a CVD system using PH3 or AsH3, yielding a maximum hole density
of about 6 x 1015 cmP3. Ghandhi et al. (1987) obtained hole densities of
2 x 1017 ~ r n -in~ epitaxial films grown on single-crystal CdTe substrates at
350°C using metalorganic chemical vapor deposition (MOCVD) growth with
AsH3 as the dopant.
150 Photovoltaic Materials
Ion-Assisted Doping ( I A D )
In the IAD process ionized and accelerated dopant beams (usually of energy
less than 100 eV) are directed onto a film as it is being grown from the vapor.
The technique has been successfully used to control the dopant incorporation
in MBE-grown Si and GaAs (Greene et al., 1985). IAD with phosphorus
impurity to produce p t y p e epitaxial CdTe films by vacuum evaporation on
single crystal substrates has been investigated as a function of ion dose, ion
energy, growth temperature, and growth rate (Sharps et al., 1990), and an
overview given by Fahrenbruch et al. (1992). A diagram of the apparatus used
is given in Fig. 5.8. The system was designed so that the CdTe effusion cell
and the ion source were simultaneously directed onto the substrate.
I ......
,________._._...--.
,_.___.._...
i-
I
: ..
Faraday Cup and Shuner
CdTe Source
Ion S ource
Fig. 5.8. Schematic diagram of the ion-assisted doping system. (Reprinted with permission
from P. Sharps et al., J . A p p l . Phys. 68, 6406 (1990). Copyright 1990 American Institute
of Physics.)
152 Photovoltaic Materials
Ion Dose ( ~ r n . ~ )
Fig. 5.9. Comparison of hole density induced in CdTe by IAD, and SIMS measurements on
incorporated phosphorus, as a function of ion dose. Squares refer to the hole density data,
and circles refer to the SIMS data. (Reprinted with permission from P. Sharps et al., J .
Appl. Phys. 68, 6406 (1990). Copyright 1990 American Institute of Physics.)
Fig. 5.10. Hole density and minority carrier diffusion length versus ion dose for IAD
pCdTe:P at 60 eV, compared to 20 eV with electron irradiation during growth. (Reprinted
with permission from D. Kim et al., J . A p p l . Phys. 7 5 , 2673 (1994). Copyright 1994
American Institute of Physics.)
154 Photovoltaic Materials
and in a medium including CdClz and/or 0 2 . The heat treatment causes CdTe
grains to regrow and become larger, and decreases the defect density at the
junction interface. The p t y p e character of CdTe films appears to be enhanced
upon annealing in 0 2 at elevated temperatures, although the exact mechanism
is not yet understood. The importance of this processing step led to the gen-
eral recipe: “The golden rule for making CdTe devices is to deposit cadmium
and tellurium as inexpensively as possible and then t o heat-treat the devices
to activate them.” (Zweibel and Barnett, 1993). The equivalent sentiment
is pictured cartoon-style in Fig. 5.11. In some cases these treatments can be
made an integral part of the film growth process itself so that a post-deposition
step is not necessary (Mitchell et al., 1985b; Basol, 1992).
Fig. 5.11. A general recipe for the preparation of high efficiency thin film CdTe solar cells.
(Reprinted with permission from B. M. Basol, International Journal of Solar Energy, 12,
25 (1992). Copyright 1992, Harwood Academic Publisher, Langhorne, PA.)
The following design rules for preparing efficient CdS/CdTe solar cells can
be summarized (Basol, 1992). (1) The CdTe films should be deposited by
any one of several suitable procedures onto a transparent CdS/TCO/glass
substrate. Here TCO stands for transparent conducting coating, e.g. SnOz or
ITO. (2) The thickness of the CdS layer is made as thin as possible to improve
the short wavelength cell response. The thickness can easily be reduced to
156 Photovoltaic Materials
below 0.2 pm using CdS deposition processes such as solution growth, elec-
troplating, and CSVT. (3) The CdTe layer is chemically treated to produce a
low-resistivity p t y p e Te-rich surface suitable for a tunneling contact with typ-
ical contact materials graphite or Ni. A p t y p e dopant such as Cu is usually
introduced at the contact/CdTe interface with an annealing step before the
contact deposition. (4) Finally a heat treatment involving 0 2 and CdC12 is an
important ingredient in the fabrication process, promoting crystal growth and
the p t y p e character of the CdTe film, as mentioned above.
The process by which an n-type, as-deposited CdTe film is converted into
p t y p e by heat treatment in air has been named the type-conversion/junction-
formation (TCJF) process. The application of the T C J F process to CdTe and
HgCdTe films produced by electrodeposition has also been described (Basol,
1984, 1988).
A two-stage deposition process has also been proposed as a possible method
for CdTe/CdS solar cell fabrication (Basol and Kapur, 1990; Bhatti et al., 1992;
Basol, 1992). This involves CdTe film growth (1) by deposition of alternating
Cd and Te layers onto a suitable substrate by such techniques as electrode-
position, vacuum evaporation, or sputtering, followed by (2) reaction of these
stacked layers to form the compound after introducing varying amounts of
CdC12 into the unreacted stacked layers to promote grain growth.
Several investigations have led to the conclusion that a CdSI-,Te, alloy is
formed at the CdTe/CdS interface due to interdiffusion that takes place during
fabrication, a process that is enhanced by the presence of 0 2 (Clemminck
et al., 1991; Birkmire et al., 1992; Mao et al., 1995; Ferekides et al., 1996).
It is believed that the formation of this interfacial layer reduces structural
defects due to lattice mismatch between the semiconductors and leads to a
lower defect density in the critical region near the interface. The extent of
interdiffusion and the value of x obtained in a particular case, depend on the
deposition temperature of CdTe and the CdC12 present during subsequent heat
treatment conditions. The tendency of 0 2 during fabrication to produce small-
grain CdS films is consistent with an enhancement of Te diffusion into CdS.
The net effect is to produce CdTe/CdS cells with higher &,c’s, fill factors, and
higher cell efficiencies.
The effect of reducing the thickness of the CdS film in polycrystalline
CdS/CdTe solar cells fabricated by vacuum evaporation, with subsequent treat-
ment with CdClz and a heat treatment in air, has been investigated (McCand-
less and Hegedus, 1991; Granata et al., 1996). Spectral response measurements
indicate that CdS and CdTe interdiffusion, corresponding to the formation of
C a d m i u m Telluride and O t h e r 11-VI Materials 157
both S-rich and Te-rich CdTel-,S, regions, increases as the CdS thickness
decreases from 1.5 to 0.12 pm. When the CdS thickness is < 0.1 pm, the
spectral response indicates the complete disappearance of a distinct CdS layer,
and the subsequent formation of a TCO/CdTel-,S, junction. In general,
the TCO/CdS,Tel-, junction is inferior to that involving CdS, so shunt cur-
rents are increased as CdS thickness is decreased. There are some exceptions
(Mitchell et al., 1985b; Nakazawa e t al., 1987), suggesting that with proper
treatment of the TCO (e.g. formation of a n insulating layer of TCO a t the
interface) & can be maintained for very thin or non-existent CdS. A related
investigation of the effect of CdS thickness on CdS/CdTe quantum efficiency
suggests that the difference between thick and thin CdS might well be as much
as 12% versus 15% in device efficiency (Granata et al., 1996).
Fig. 5.12. Measured band gaps of as-deposited CdTel-,S, thin films as a function of x, as
reported in this work (X), Bonnet 1970 (triangles), Hill and Richardson 1973 (diamonds),
Ohata et al., 1973 (circles), and Al-Ani et al., 1993 (squares). (Reprinted with permission
from D. G . Jensen et al., in T h i n Films f o r Photovoltaic and Related Device Applications,
eds. D. Ginley, A. Catalano, H. W. Schock, C. Eberspacher, T. M. Peterson and T. Wada,
Materials Research Society Proc. 426, Pittsburgh, PA (1996a), p. 327.)
the solubility limit, gave an efficiency of 10.8% and showed cell characteristics
very much like those of a conventional CdTe/CdS cell where the alloy is formed
by cell processing.
Transmission electron microscopy (TEM) was used to investigate the mi-
crostructures and interfaces of thin film CdTe/CdS/ITO/glass photovoltaic de-
vices, particularly as a result of annealing in air after prior exposure to CdCl2
(Loginov et al., 1996). The cells used were prepared by vacuum evaporation of
CdS onto an ITO/glass substrate, followed by CSVT deposition of CdTe. It
is concluded that a CdC12 treatment reduces the influence of interfacial states
on the current transport mechanism (changing from tunneling to thermally
activated), causes grain growth in the CdS and CdTe, and decreases the den-
sity of stacking faults and dislocations in the CdTe. It also introduces C1-rich
precipitates into the lattice, but these have not been observed in the region
near the interface, and their effect on device performance remains unknown.
TEM indicates that the CdTe grains grow epitaxially on the CdS grains
during deposition and that heat treatment causes substantial grain growth in
the CdTe, such that the epitaxial relationship is lost and one CdTe grain may
bridge several differently oriented CdS grains (Al-Jassim et al., 1993).
An investigation of the micro through nanostructure properties of polycrys-
talline CdTe using atomic force microscopy (AFM) has shown correlations with
processing and electronic structures for several standard deposition techniques
such as vacuum evaporation, CSVT, and sputtering (Levi et al., 1994). Corre-
lations between nanostructure and optical properties indicate that CdC12 heat
treatment passivates defects. Except for CSVT-deposited films, the CdCl2
treatment also enhances grain size. Diffusion of sulfur across the CdTe/CdS
interface during heat treatment has been verified.
The CdClz anneal of CdTe/CdS solar cells usually consists of coating the
CdTe surface with a saturated solution of CdC12 in methanol and heat treating
the film in air at 400°C to 450°C for 10 to 30 minutes. An alternate approach
using a CdClz vapor treatment has been developed that results in uniform
large grain films, preferred (111) crystallographic texture, less chloride surface
residue, less S diffusion into the CdTe, and more reproducible current-voltage
performance than is obtained with other CdC12 treatments (Zhou et al., 1994;
Birkmire et al., 1995; McCandless et al., 1996). In the vapor treatment, sam-
ples were annealed at 425°C for 20 minutes in air with a partial pressure of
CdC12 of 0.04 Torr. Vapor treated films do not require a rinse step prior to
device completion, unlike samples that are CdCl2:methanol treated. Solar cells
prepared with the CdClz vapor treatment were comparable to the best CdTe
Cadmium Telluride and Other II-VI Materials 159
cells: & = 0.83 V, J,, = 19.6 mA/cm2, and f f = 0.65. Similar microstruc-
tural changes were observed for HC1 vapor treatment of films (Qu et al., 1996).
Interdiffusion between CdS and CdTe during cell preparation has been
deliberately limited by a process in which the CdS film on a glass substrate
is treated with ZnC12 and CdClz to reduce the reactivity of the CdS when
CdTe is subsequently deposited on top of the CdS by close-spaced sublimation
(Romeo et al., 1997). Resulting cells showed &, = 0.845 V, J,, = 23 mA/cm2,
f f = 0.71, and efficiency of 13.8%.
The need to evaluate the uniformity of the microstructure of CdS/CdTe
large-area solar cells has led to application of the non-destructive testing
method of time-resolved photoluminescence to map recombination lifetimes
in such cells (Ahrenkiel, 1992; Ahrenkiel et al., 1994). Typical lifetime profiles
indicated spatial variation by factors of two to three across 1 cm dimensions,
and that the open-circuit voltage of the cells was controlled by the area with
the minimum lifetime.
One possible solution to the problems posed by difficulties with doping and
contacts on p C d T e has been proposed by Meyers (1989). The solar cell consists
of a p i - n heterojunction made up of n-CdSli-CdTelpZnTelNi, as shown
in Fig. 5.13. The n-CdS layer is deposited on SnOz-coated glass, the CdTe
r
V
-SnO, - 0.4 pm
-CdS - 0.15 prn
-CdTe - 2 0 pm
t -2nTe-0.1
- - prn
Ni 0.05 prn
Fig. 5.13. Solar cell structure for an 11%CdTe n - i p solar cell. At 1 sun,
# = 0.68 V,
J ~ ~
Jsc = 20.5 mA/cmZ, fill factor = 0.625. (Reprinted with permission from P. V. Meyers,
Photovoltaics Advanced Research and Development (1989), 9th Annual Review Meeting,
SERI/CP-213-3495, p. 19. Copyright 1989, SERI, Golden, CO.)
160 Photovoltaic Materials
layer is deposited by electrodeposition on the CdS, and the final ZnTe layer is
deposited on the CdTe. Work on this kind of cell has been extended by Nouhi
et al. (1989) with the use of CdTe films deposited by MOCVD to achieve
small-area cells with an efficiency over 9%.
SIS-type cells have also been prepared with polycrystalline ZnCdTe or
CdMnTe (Rohatgi et al., 1988, 1989) as the i-region material; these devices
have a lower 4octhan similar devices using CdTe alone.
o ON GLASS
\
8\, LIGHT ?
0
10' - \
L I I I I
370 470 5 70
670
INDIUM SOURCE TEMPERATURE ("C)
Fig. 5.14. Dark and light resistivity of CdTe:In films grown on glass and on single crystal
BaFz by HWVE, as a function of the indium source temperature. Films on glass in the
dark ( 0 ) and in the light (0);films on BaF2 grown with Ts,,b = 480OC in the dark (A) and
in the light (A); films on kaFz grown with T S U b = 450OC in the dark (D) and in the light
(0). (Reprinted with permission from W. Huber et al., J. A p p l . Phys. 54, 4038 (1983).
Copyright 1983 American Institute of Physics.)
Cadmium Telluride and Other II-VI Materials 161
ON GLASS
-104
-
0
-lo3 5
.
n
-=L
E
0
-lo2 >
c2
-
m
0
-10' =
RESISTIVITY p ( R cm)
Fig. 5.15. Variation of the electron density and mobility under optical excitation as a function
of the resistivity in the light for CdTe films deposited by HWVE (a) on glass substrates, and
(b) on a BaFz substrate. (Reprinted with permission from W . Huber et al., J . Appl. Phys.
54, 4038 (1983). Copyright 1983 American Institute of Physics.)
substrates (Huber e t al., 1983). Resistivity and Hall effect measurements could
be successfully interpreted in terms of the grain boundary model first proposed
by Set0 (1975), which is a one-dimensional model including the contribution of
the bulk resistivity of the undepleted portion of the grain, and carrier transport
across the barrier described by thermionic-diffusion theory (Sze, 1969).
The dependence of the dark and light resistivity of the two types of films
on the temperature of a n indium source used for doping is shown in Fig. 5.14,
and the variation of the electron density and mobility under photoexcitation,
measured by photo-Hall effect, with the light resistivity for the two types of
films is shown in Fig. 5.15. The dark resistivity of CdTe films deposited on
162 Photovoltaic Materials
Fig. 5.16. Grain-boundary conductivity as a function of temperature in the dark and for
several different illumination intensities (10 = 60 mW/cm2 white light) for the grain bound-
ary in a p t y p e CdTe bicrystal. The inset shows the variation of the activation energy with
light intensity at low temperatures. (Reprinted with permission from T. P. Thorpe et al., J .
A p p l . Phys. 60, 3622 (1986). Copyright 1986 American Institute of Physics.)
Cadmium Telluride and Other II-VI Materials 163
wz --After H2 Diffusion
Fig. 5.17. Density of grain-boundary interface states as a function of energy in the forbidden
gap of CdTe, for a particular p t y p e CdTe bicrystal, before any passivation, after passivation
by Li diffusion and after passivation by hydrogen heat treatment. (Reprinted with permission
from T. P. Thorpe et al., J . Appl. Phys. 60,3622 (1986). Copyright 1986 American Institute
of Physics.)
164 Photovoltaic Materials
conduction band (Thorpe et al. 1986). Figure 5.17 shows a typical variation
of grain boundary state density without any passivation, and the effects of
Li diffusion and H2 heat-treatment passivation. The passivation effects are
striking, but unfortunately are not stable for more than a few weeks. Similar
experiments with grain boundaries in n-type CdTe showed that in this case
passivation could be temporarily achieved by heating in air, not in H2.
10:
CdS- -CdTe
10'
I
(A)
16'
lGf
Id' I I I 1 I I
Fig. 5.18. Electron-beam-induced current a s a function of beam position for the 7.9% n-
CdSlpCdTe cell. The calculated values of the minority carrier diffusion lengths are 0.40 pm
in pCdTe, and 0.43 p m for n-CdS. (Reprinted with permission from K. W. Mitchell et al.,
J . A p p l . Phys. 48, 4365 (1977). Copyright 1977 American Institute of Physics.)
C a d m i u m Telluride and Other II-VI Materials 165
(Mitchell et al., 1977c), and an analysis of the fill-factor in such cells was car-
ried out (Mitchell et d., 1977a). Sputtered indium-tin-oxide (ITO) was used as
a transparent electrode to the CdS, and glycerol was used as a n anti-reflection
coating. Two CdTe surface treatments were compared (mechanical polish-
ing, and mechanical polishing followed by a bromine-in-methanol (MB) etch),
and the effects of different CdTe and CdS resistivity were investigated. The
MB-etched cells were significantly better than the mechanically polished cells,
giving a saturation current J , some one to two orders of magnitude smaller,
and a quantum efficiency significantly larger. A typical spectral response of
the cells is shown in Fig. 1.13in Chapter 1 of this book, and a plot of electron-
beam-induced current (EBIC) for the 7.9% cell is given in Fig. 5.18.
A variation of vacuum evaporation called hot-wall vacuum evaporation
(HWVE) is illustrated in Fig. 5.19, similar to that used by Lopez-Otero (1978)
and Lopez-Otero and Huber (1979). The apparatus fits inside a standard
vacuum system, and consists of a heated quartz liner with three independent
furnaces that allow an independent control of the CdTe, dopant, and Cd or
Te source materials. The substrate is placed in a fourth furnace on top of
the quartz liner. The use of a heated wall confining the region of evaporation
almost eliminates waste of material and provides an environment in which the
deposited films are closer to equilibrium conditions. Many of the phenomena
encountered in depositing CdTe films are similar using CSVT or HWVE. In-
vestigation of grain boundary effects using CdTe films deposited by HWVE is
described above (Huber et al., 1983).
Polycrystalline CdTe films were deposited by HWVE on graphite sub-
strates, and single-crystal CdTe films were deposited by HWVE on single-
crystal pCdTe:P substrates (Fortmann et d., 1987). CdS/CdTe heterojunc-
tions were formed by the vacuum evaporation of CdS. It was observed that the
q50c was a maximum at about 0.80 V for both types of cells when the carrier
densities in the n-CdS and p C d T e were approximately equal. The effects were
successfully modeled in terms of an unidentified level lying 0.45 eV below the
bottom of the conduction band in CdTe.
Difficulties in doping CdTe films using a doped source, such as are described
below for CSVT film deposition (Anthony et al., 1985), were also encountered
with attempts to dope CdTe using HWVE (Huber et al., 1983; Anthony et al.,
1984; Bube et al., 1984). Attempts were made to dope CdTe films by co-
evaporation of such dopants as As, Sb, Na, and Ag, but all attempts to dope
p t y p e CdTe films during HWVE deposition were unsuccessful. Likely causes
include: (1) small sticking coefficients for the dopants on the growing surface
166 Photovoltaic Materials
strale Carrier
F n 1 Zone
(Substrate) Shutter Guide
huller (Ouarlz)
Secondzone
(Source) CdTe Source Material
Third Zone
I
Radlalton Shields
Heater Wound on
tainless Steel Core
Fourth Zone Third Source (Cd or Tel
Thermal Links lo H 0
Cooled Baseplate
Fig. 5.19. Schematic diagram of the hot-wall vacuum evaporation (HWVE) apparatus used
to grow CdTe thin films. The chamber pictured is enclosed in a vacuum system. (Reprinted
with permission from R. H. Bube et al., I E E E Transactions on Electron Devices ED-31,
528 (1984). Copyright 1984, IEEE, NY.)
Fig. 5.20. Equiefficiency curves for the pCdTe/n-CdS heterojunction system as a func-
tion of quantum efficiency and reverse saturation current density. Calculated for negligible
reflection, series resistance and area utilization losses. (Reprinted with permission from
A. L. Fahrenbruch et a!., Appl. Phys. Lett. 25, 605 (1974). Copyright 1974 American
Institute of Physics.)
168 Photovoltaic Materials
QUARTZ-HALOGEN
Fig. 5.21. Basic features of the close-spaced vapor transport (CSVT) system used t o grow
CdTe thin films. (Reprinted with permission from T. C. Anthony et al., J . Vac. Sci.
Technology, A2, 1296 (1984). Copyright 1984, American Institute of Physics.)
consequences of this early research was the realization that the reverse satu-
ration current density J , was about lo3 times larger than the lowest values
observed for junctions prepared by other means, resulting in values of open-
circuit voltage between only 0.5 and 0.6 V instead of the highest observed value
of about 0.9 V. Figure 5.20 shows the dependence of cell efficiency on quan-
tum efficiency and J , for pCdTe/n-CdS heterojunctions; for these early cells
prepared by CSVT, the quantum efficiency was about 0.85 and J , was about
3x A/cm2. Use of CSVT for the growth of CdTe films has also been
reported by Mimila-Arroyo et al. (1977, 1979) and Tyan and Perez-Albuerne
(1982).
A diagram of the CSVT apparatus used by Anthony et al. (1984) is given
in Fig. 5.21. The source and substrate susceptors, as well as the mask, are ma-
chined from purified, high-density graphite. Ceramic-sheathed chromel-alumel
thermocouples monitor the temperatures of the graphite blocks that are heated
independently by two 650 W quartz-halogen lamps. A stainless-steel cooling
tube positioned within the substrate block enables lower substrate tempera-
tures to be achieved and permits rapid cooling of the substrate. In addition,
C a d m i u m Telluride and Other IZ-VI Materials 169
the spring action of the cooling tube ensures good thermal contact between
the substrate and the substrate block. A perforated] cylindrical quartz spacer
maintains a separation of about 1 mm between the source and substrate. The
graphite source block is supported by a quartz plate and the entire chamber
is enclosed within a quartz tube. It was concluded that the growth of CdTe
films by CSVT in inert ambients at atmospheric pressure is diffusion limited.
Important parameters in the diffusion theory are the spacing, ambient gas,
ambient gas pressure, and source temperature, each offering independent con-
trol over the growth rate. Approximately equal growth rates in He and in H2
indicated the relative ineffectiveness of Hz as a transport agent for CdTe.
In a subsequent investigation of the electrical properties of CdTe films and
junctions] polycrystalline films of p C d T e were deposited by CSVT on graphite
Fig. 5.22. Uncompensated ionized acceptor density in CdTe films deposited by CSVT as
a function of the graphite substrate temperature, using a source temperature of 653OC
in hydrogen at 1 atm, for several different sources doped as indicated. (Reprinted with
permission from T. C. Anthony et al., J . Appl. Phys. 57, 400 (1985). Copyright 1985
American Institute of Physics.)
170 Photovoltaic Materials
(1982), and of Chu et al. (1987), it appears that the inverted configuration
(CdTe on CdS, rather than CdS on CdTe) allows fine tuning of the junction
parameters to achieve higher cell performance.
Although almost all workers have found the CdS layer to be necessary
for high &, it is possible to make cells without it. Mitchell et al. (1985b)
deposited polycrystalline CdTe directly on tin-oxide coated glass by CSVT.
These 4 cm2 cells produced doc= 0.66 V, J,, = 28.1 mA/cm2, and an efficiency
of 10.5%.
The effect of various deposition parameters on CdS films deposited by
CSVT on Corning 7059 glass, and the subsequently formed CdTe/CdS junc-
tion, finally resulting in an efficiency of 15.8%, has been investigated in some
detail using measurements of SEM, photoluminescence, current versus voltage
versus temperature, spectral response, and capacitance versus voltage mea-
surements (Ferekides et d., 1993, 1996). It was found that the presence of 0 2
during deposition of CdS produced smaller and denser grains in the CSVT-
deposited CdS films, resulted in a considerable decrease in the deposition rate
that could be compensated for by raising the source temperature, and resulted
in improved solar cell performance. Measurements indicate that this is the
result of the addition of 0 2 forming pinhole free, thin (70-80 nm) CdS films,
eliminating the possibility of the subsequently deposited CdTe coming into
direct contact with the SnO2/glass substrate. The inclusion of 0 2 in the de-
position also appears to enhance the formation of a CdSI-,T, region due to
Te diffusion into the CdS films, which shifts the electrical junction away from
the metallurgical interface, as discussed above in Sec. 5.2.
There still appears to be incomplete agreement on the necessity for 0 2 ,
however. CdS/CdTe solar cells have been prepared by CSVT with an effi-
ciency of 13% without the inclusion of any oxygen in a He-only ambient by
using thicker CdTe films to minimize pinholes, and a stronger CdClz anneal
to increase interdiffusion a t the CdS/CdTe interface. (Rose et al., 1996a,b).
By modifying the source and substrate temperature profiles used during
CSVT deposition, it has been shown that it is possible to grow large-grain
material, fill pinholes, improve device &,-, and achieve a n efficiency of 12% (Li
et al., 1996).
HEATER s
Fig. 5.23. Schematic diagram of a spray-pyrolysis (SP) apparatus (Reprinted with permission
from Y . Ma, Ph.D. Dissertation, Stanford University (1977).)
Fig. 5.24. Spectral response of quantum efficiency for three cells of n-CdSlpCdTe prepared
by SP deposition of CdS onto single-crystal pCdTe. Sample M-2 used one CdTe boule,
samples M-8 and M-14 used a second boule. CdTe substrate temperature was 45OOC for
M-2 and 400OC for M-8 and M-14. HZ heat treatment was for 5 min at 400OC for M-2 and
M-8, and for 5 min at 42OOC for M-14. Solar efficiency of M-2 was 6.5% (uncorrected for
reflection losses). (Reprinted with permission from Y . Y . Ma et al., A p p l . P h y s . Lett. 30,
422 (1977). Copyright 1977 American Institute of Physics.)
does not vary appreciably with x; in the Zn,Cdl-,S system, however, the
resistivity increases almost exponentially from about 10 ohm-cm for 5 = 0 to
10" ohm-cm for x = 1.0, due both to a n increase in the ionization energy
of chlorine donors and to compensation of these donors with increase in x
(Chynoweth and Bube, 1980).
A series of photovoltaic heterojunctions of the type n-Zno.lCdo.gS/pCdTe
have been prepared by SP of ZnCdS films on single-crystal CdTe with variation
of the CdTe substrate temperature and of the postdeposition heat-treatment
temperature in hydrogen (S.-Y. Yin et al., 1978). Results indicated that the
heterojunction system itself could provide &c = 0.8 V with an efficiency of 8%
without an antireflection coating if the technological problems associated with
the contacts to the ZnCdS film could be resolved.
A detailed investigation was made of the electrical and photovoltaic pro-
perties of heterojunctions prepared by SP deposition of thin ZnO films on
single-crystal p t y p e CdTe:P (Aranovich et al., 1979, 1980), using the sub-
strate temperature and the postdeposition temperature for annealing in H2 as
the principal experimental variables. In spite of a large 28% lattice mismatch
between ZnO and CdTe, high quantum efficiencies were achieved with values
greater than 90% if a reverse bias was applied, the spectral response is that ex-
pected for a heterojunction, and cells with efficiency up to 8.8% were obtained
referred to the active area of the cell. A thermally-assisted tunneling model
(Padovani and Stratton, 1966) in which bulk and interface deep traps control
the forward characteristics provides a good correlation with experimental data.
Spray pyrolysis has been used for the production of relatively large area
SnO,/CdS/CdTe/Ni/Sn solar cells, with small area samples showing efficien-
cies greater than 12% (Albright et al., 1988, 1989; Jordan et al., 1988).
It has been shown that CdTe films and a CdS/CdTe heterojunction can
be deposited on stainless-steel substrates using spray pyrolysis, although con-
siderable refinements would be required for a viable solar cell (Berry et al.,
1990).
It has been proposed that high-quality CdTe films can be produced by
SP using nanoparticle precursors making possible materials with improved
smoothness, density, and a lower processing temperature (Schulz et al., 1996).
CdI2 was reacted with NazTe in a methanol solvent, resulting in the formation
of soluble NaI and insoluble CdTe nanoparticles. CdTe thin films were pro-
duced by spray depositing the nanoparticle colloids (2.5-7.5 nm diameter) onto
substrates at 280-440°C with no further thermal treatment. A layer-by-layer
film growth mechanism was proposed for such a one-step spray deposition of
nanoparticle precursors.
Cadmium Telluride and Other It-VI Materials 175
Electrodeposition (ED)
It was discovered early that reasonably efficient CdTe solar cells could be fabri-
cated using the relatively simple techniques of electrodeposition (Basol, 1988;
Meyers, 1988). Panicker et al. (1978) laid the foundation for this by showing
the feasibility of electroplating CdTe thin films. Fulop et al. (1982) produced
an MIS cell involving an n-CdTe film deposited by cathodic electrodeposi-
tion from an acidic aqueous solution of Cd- and Te-containing ions, and a n
insulating oxide layer obtained by annealing in air after deposition. The elec-
trodeposition technique has been developed by Basol et al. (1985), Meyers
(1989), Das and Morris (1992), Woodcock et al. (1993, 1994), and Paulson
et al. (1994). Lincot et al. (1995) have investigated the epitaxial electrode-
position of CdTe films on InP; the presence of a thin layer of CdS, formed by
chemical bath deposition on the InP markedly improved the epitaxial growth
of the CdTe.
An investigation of the formation of CdS/CdTe solar cells using the tech-
niques of electrodeposition resulted in a cell with an efficiency of 11.0% (Kim
et al., 1994a). CdS layers were deposited by chemical bath deposition on
Sn02-coated glass substrates, using cadmium acetate, ammonium acetate and
ammonium hydroxide as described by Chu et al. (1992). A CdCl2 coating
was applied to the surface of the CdS, the sample was annealed at 450°C for
50 min in N2, and then was rinsed in hot DI water and dried in Nz. The CdTe
deposition was done in an electroplating system with Cd and Te anodes, a
Ag/AgCl reference electrode, and the sample substrate as the cathode; after
deposition the CdTe samples were rinsed in hot DI water and dried in NZ. A
CdC12 coating was applied to the CdTe surfaces, samples were annealed at
410°C for 45 min in air, cooled, rinsed and dried. Contacts to the CdTe were
made by evaporating Au through masks to form dot cells (0.033 cm2) after
etching with Br-MeOH. Contacts to the CdS were made by indium solder and
silver paint. EBIC measurements indicated a maximum inside the CdTe layer,
suggesting a buried homojunction.
C a d m i u m Telluride and Other IZ-VI Materials 177
Kim et al. (1994b) also investigated the effect of annealing on the mi-
crostructure and stress state of electrodeposited CdTe, involving recrystalliza-
tion and grain growth. Additional discussion of stress testing of cells has been
given by Meyers and Phillips (1996).
Using a similar approach, Kampmann et al. (1995) carried out a detailed
investigation of the influence of the electrodeposition potential on the optical,
photoelectrochemical, and structural properties of CdTe. An efficiency of over
12% has been reported for thin film cells fabricated using electrodeposition
(Song et al., 1996).
CdS films of device quality were deposited by electrodeposition by electrore-
duction of Cd and S at the substrate in a bath of CdClz and sodium thiosulfate
(McCandless et al., 1995). These films were similar to those deposited by vac-
uum evaporation, respond similarly to air heat treatments, and can be used in
the fabrication of CdTe/CdS solar cells with efficiencies comparable to those
made with evaporated CdS. Electrodeposition provides a fabrication method
using a low temperature (85OC), very low deposition current densities, and
high utilization of constituent materials.
ZnO has been of interest as a transparent n-type window in solar cells.
ZnO films of good quality have been prepared by cathodic electrodeposition
from aqueous solutions at low deposition temperatures ( 5 80°C) (Peulon and
Lincot, 1995, 1996). These films have high transparency in the visible corre-
sponding to a band gap of about 3.4 eV, and high n-type doping is possible.
The structure of the films can be varied by control of the deposition conditions
from dense films to open-structured layers of columnar crystals.
Sputtering ( S T )
The optical and electrical properties of indium-tin oxide (ITO) sputtered on
glass have been investigated to determine the conditions for maximum conduc-
tivity and transmissivity (Haines and Bube, 1978).
When an ITO/CdTe junction is formed by rf sputtering of the I T 0 onto
single crystal CdTe, a 1-pm thick surface layer of the p t y p e CdTe is converted
to n-type as a consequence of the changes in the defect structure induced by
ion bombardment and temperature, producing a buried homojunction of the
type n-ITO/n-CdTelpCdTe with solar efficiencies up to 8% and & = 0.82 V
(Courreges et al., 1980). A proposed energy band diagram is given in Fig. 5.25.
Unfortunately the n-CdTe layer is unstable and shows degradation in pho-
tovoltaic parameters as the homojunction is converted to a heterojunction
at room temperature. Related investigation of the properties of In/CdTe
178 Photovoltaic Materials
x = 4.28eV
r
I .....
,
a = 4.5eV
I
E,= 1 . W V
E f l
VB
Fig. 5.25. Proposed energy-band diagram for an ITO/CdTe junction produced by rf sput-
tering of I T 0 onto the surface of a p t y p e CdTe:P crystal, showing the n-type CdTe layer
formed by thermal effects during sputtering. (Reprinted with permission from F. G. Cour-
reges et al., J . A p p l . Phys. 51,2175 (1980). Copyright 1980 American Institute of Physics.)
(Lincot and Vedel, 1991; Lincot and Ortega-Borges, 1992; Ortega-Borges and
Lincot, 1993). They conclude that film growth proceeds in two consecutive
steps, the formation of a dense compact inner layer followed by the formation
of a porous less adherent layer. A simple columnar growth model is shown
to account reasonably well for the experimental results. CBD allows the de-
position of very thin films of the order of a few nanometers in the presence
of excess thiourea, a possibility that can be quite important in seeking high-
efficiency solar cells with CdS window layers. The process has been studied
further through the use of high-resolution transmission electron microscopy,
which indicates that the growth in CdS probably proceeds by way of an atomic
mechanism, rather than by cluster coagulation as is indicated for CBD growth
of ZnS and CdSe (Lincot et al., 1993; Froment et al., 1995). The growth of
CBD CdS thin films on SnO2 and glass substrates has been analyzed using
atomic force microscopy (AFM) (Moutinho e t al., 1996). CBD CdS films have
been optically and electrically characterized (Ozsan e t al., 1994). The band
gap of these films depends strongly on the deposition conditions, increasing
with an increase in deposition rate. The films are all n-type with doping den-
sities aground 1017 cmP3 under illumination without intentional doping in the
bath.
CBD CdS films have been recently investigated in the formation of CdS/
InP, CdS/CdTe, and CdS/CuInSea junctions. Epitaxial growth of hex-CdS on
(111)*InP single crystals has been achieved using CBD from cadmium amm-
onia-thiourea aqueous solutions (Lincot et al., 1994; Froment et al., 1995), and
a solar cell based on a p I n P single crystal with InP/CBD-CdS/In203 structure
gave an efficiency of 17.8% (Saito et al., 1994). Epitaxial growth of cub-CdS on
(100) InP was achieved by CBD, probably because of lattice matching between
the CdS (100) plane and the InP (100) plane (Cortes et al., 1996).
The interaction between CBD deposited CdS films and subsequently elec-
trodeposited CdTe films in the formation of CdS/CdTe junctions has been
investigated (Ozsan et al., 1995). Results indicated that the as-deposited cub-
CdS CBD films with small grain size and high density of defects are beneficial
for the recrystallization in the cell structure during the subsequent heat treat-
ment.
CBD CdS films were successfully used in the fabrication of ZnO/CdS/
CuInSe2 solar cells, where it was found that the final cell performance depends
on the specific Cd-salt used in the CBD CdS film deposition process: High
efficiency was obtained for CdIB and CdS04 (Mauch et al., 1991a,b). Further
investigation indicated that the CBD reaction is not limited to the abrupt de-
position of a CdS layer on CuInSez, but involves chemical modifications at the
Cadmium Telluride and Other 11-VI Materials 181
by CSVT (Peters et al., 1988a,b) and by MBE (Rohatgi et al., 1988, 1989),
ZnSel-,Te, using LPE (Koval et al., 1991), and Cdl-,Mn,Te using deposition
by MOCVD (Nouhi et al., 1989; Rohatgi et al., 1988, 1989).
Zn,Cdl-,Te films with Zn concentrations as high as x = 0.45 were de-
posited onto graphite substrates using CSVT (Peters et al., 1988a). Values of
x in the deposited films equal to that in the source could be achieved either
by deposition with a small temperature difference (2-8°C) between source and
substrate at atmospheric pressure of an ambient gas, or by deposition at low
ambient pressures (0.3 Torr) with any desired temperature difference between
source and substrate. For films deposited at a substrate temperature of 610°C,
p = 5 x 1015 cmP3 as determined from capacitance versus voltage measure-
ments on In/ZnCdTe Schottky barriers. No change in this hole density was
found with variations in Zn concentration.
A series of CdS/Zn,Cdl-,Te heterojunctions were prepared by vacuum
evaporation of CdS onto single crystal p t y p e Zn,Cdl-,Te substrates with
x = 0, 0.1, and 0.3 (Peters et al., 1988b). These junctions showed a max-
imum q5,, when the Zn,Cdl-,Te surface was made stoichiometric by heat
treatment in Hz before CdS deposition, a maximum q5,, for any value of x
for CdS deposition conditions yielding approximately equal carrier densities
in CdS and Zn,Cdl-,Te, and an increase in J, with increasing x. The best
junction for x = 0.3 had a q5,, of only 0.36 V. The junction transport process
resulting in relatively high values of J , (2.6 x A/cm2 in the best cell with
x = 0.3), could be modeled by thermally assisted tunneling of holes from the
Zno.3Cdo.i.Te to the interface of the junction followed by recombination there
with electrons from the CdS, provided that a high density of interface states
was assumed. It was suggested that the large values of J , were associated with
lattice distortions resulting from the solid solution formation.
Thin films of Hg,Znl-,Te over the range from 5 = 0 t o x = 0.4 have
been deposited by electrodeposition (Thero and Singh, 1996). These films
show optical absorption coefficients of the order of lo4 cm-', indicating their
possible suitability for use as absorber materials in thin film solar cells.
The processing steps involving TCJF, described in Sec. 5.4, necessary for
a solar cell involving electrodeposited HgCdTe have been discussed (Basol and
Tseng, 1986; Basol, 1988). A completed CdS/Hgo.lCdo.gTe cell was charac-
terized by q5,, = 0.62 V, J,, = 27.03 mA/cm2, f f = 0.63 and efficiency =
10.6%. Compared to a similarly electrodeposited CdS/CdTe cell (band gap of
CdTe = 1.47 eV, compared to that of Hgo.lCdo.gTe = 1.30 eV) with the same
efficiency, the CdS/HgCdTe device shows a 27% larger J,, and a 14% smaller
4oc.
Cadmium Telluride and Other II-VI Materials 183
4 Ix (Relative scale)
1.0
0.1
0.1
0.4
0.2
0
1.4 1.1 1.1 2.0 2.2 2.4 2.1 2.1 3.0 3.2
hv ( e V )
Fig. 5.26. Short-circuit spectral response curves for four 11-VI compound solar cells prepared
by CSVT. Both M-1 and M-21 are n-ZnSelpCdTe cells, but their spectral responses are
quite different. The response of M-1 resembles that of n-CdSelpCdTe cell N-3, suggesting
CdSe formation at the junction in M-1. The low-energy spectral cutoff for M-21 resembles
that of pZnTe/n-ZnSe cell P4, suggesting that in this case ZnTe is formed at the junction.
(Reprinted with permission from F. Buch e t al., J . A p p l . Phys. 48, 1596 (1977). Copyright
1977 American Institute of Physics.)
CHAPTER 6
CuGoS2 - - Cd Sipz
- Z n Si Pz
Ag I n s 2
A9 GOS C p
-- + L n SnAsZ
CUGOSCZ - CdGrb
-2n S n P2
AqGoTe2 -
A9 I n Sa2- - -
-eC d Sn Pz
CU GoTe2-
Co n e2-
-- Y! n Gc A s
$?
A9 n e2=
CuInTc2
=-1.0-
- Z n SnAs2
Fig. 6.1. Lowest energy gaps of I-111-VI2 and II-IV-V2 semiconductors in the range of
interest for solar photovoltaic conversion (Reprinted from S. Wagner and P. M. Bridenbaugh,
J. Crystal Growth 39,151 (1977) with kind permission of Elsevier Science - NL, Sara Bur-
gerhartstraat 25, 1055 KV Amsterdam, The Netherlands. Copyright 1977 Elsevier Science.)
184
Copper Indium Diselenide and Other I-III-VI Materials 185
Such a ternary compound represents the first step to more complicated solid
solutions with more than three different atoms, so that both the desired band
gap and the proper heterojunction lattice match could in principle be obtained
to provide maximum efficiency. Research on quaternary and pentenary systems
has been undertaken (Antypas and Moon, 1974; Loferski e t al., 1981) but to
date these much more complex materials have not met the stringent demands of
reproducibility and operating lifetime required for a solar cell material. Some
twenty-three ternary compounds of type I-111-VI2 and 11-1v-V~ with band
gaps in the range of interest for solar cells are summarized in Fig. 6.1.
In 1974 it was shown that a solar cell efficiency of 12010, a high value for
cells of any type at that date, could be obtained with a CdS/CuInSe2 het-
erojunction made by vacuum evaporation of CdS onto single crystal p t y p e
CuInSea, previously Se-annealed to increase its p t y p e conductivity (Wagner
et al., 1974; Shay et al., 1975).
Fig. 6.2. Current versus voltage curves under illumination by 100 mW cmP2 by tungsten-
halogen for early solar cells: all-thin-film heterojunctions of CdSjCuInSe2 and CdS/CuInSz,
and all-thin-film homojunctions of CuInSez and CuInS2. (Reprinted with permission from
L. L. Kazmerski, Univ. of Maine, Orono, Second Quarter Report t o NSF-RANN and ERDA,
NSF/AER 75-19576/PR/77/2, p. 13 (1977).)
186 Photovoltaic Materials
Fig. 6.3. Spectral response of quantum efficiency for early solar cells: all-thin-film hetero-
junctions of CdS/CuInSez and CdS/CuInS2, and all-thin-film homojunctions of CuInSe2
and CuInS2. (Reprinted with permission from L. L. Kazmerski, Univ. of Maine, Orono,
Second Quarter Report t o NSF-RANN and ERDA, NSF/AER 75-19576/PR/77/2, p. 16
(1977).)
Copper Indium Diselenide and Other I-III-VI Materials 187
2-3pm
/k AI grid contact
Glass. alumina. a r
foil substrate
Fig. 6.4. Solar cell structure for a 9.6% large area CuInSez solar cell. At 1 sun, = 0.43 V,
.Isc= 35.3 mA/cm2, and fill factor = 0.63. (Reprinted with permission from W. E. Devaney
and R. A. Mickelsen, Solar Cells 24, 19 (1988). Copyright 1988, Elsevier Sequoia S.A.,
Lausanne, Switzerland.)
Fig. 6.5. Solar cell structure for the A R C 0 (now Siemens) ZnO/CdS/CuInSez solar cell. For
simulated sunlight of 100 mW/cm2, &, = 0.36-0.42 V, J,, = 35-40 mA/cm2, ff = 0.5-
0.65, and cell efficiency = 7-9%. (Reprinted with permission from J. B. Yoo, et al., J . Appl.
Phys. 68, 4694 (1990). Copyright 1990 American Institute of Physics.)
188 Photovoltaic Materials
CuInSez as the absorbing material. Table 6.1 provides data and references for
specific solar cell parameters as of 1987. A reported CdS/CuInTea cell showed
no photovoltaic effect.
Overall better performance in CdS/CuInSez cells was achieved using the
backwall configuration shown in Figs. 6.4 and 6.5. Three reasons were ad-
vanced for this: (1) the lower deposition temperature for CdS as compared to
CuInSe2 minimizes diffusion between the two materials, (2) the CdS or CdZnS
window layer protects the CuInSez absorber layer from interaction with the
surface environment, and (3) a heat treatment in oxygen that proves critical
for optimizing cell performance is effective only if the CdS or CdZnS is on top.
C. CuInl-,Ga,Se2 Cells
ZnCdS/ 0.51 30.4 10.2 Chen et al., 1987
CuIno.77Gao.z3Sez
ZnCdS/CuGaSe2 0.85 11.6 5.8 Dimmler et al., 1987
ZnCdS/ 0.52 15.6 2.9 Dimmler et al., 1987
CuIn0.44Gao.56Sez
CdS/CuGaSez 0.58 9.6 2.6 Noufi et al., 1986
Copper Indium Diselenide and Other I-III-VI Materials 189
k
z
w
u
a
w
n
G
u
z
w
0
u.
u.
w
Fig. 6.6. NREL solar cell efficiency a s a function of year for all thin-film CuInSe2 cells.
(Data from R. F. Service 1996.)
The increase in efficiency of all thin-film CuInSea solar cells over the years
up to the present is shown in Fig. 6.6. Reviews of the progress have been given
by Meakin (1985), Rockett and Birkmire (1991), Basol (1992), and Tarrant
and Gay (1995). By 1996, several improvements developed over the intervening
years, including increasing the crystallite size by varying the concentration of
copper as the film is deposited with a copper-rich layer a t the bottom, and the
inclusion of Ga to make a larger band gap CuInl-,Ga,Sez (CIGS) material
graded with larger values of z near the bottom of the film to aid in electron
collection, have led to a 0.4 cm2 cell with an efficiency of 17.7% (J. Tuttle
190 Photovoltaic Materials
et al., 1996), marking the high point on Fig. 6.6. At the same time a CIGS
cell with 100 times larger area has shown a n efficiency of 13.9% (Schock et al.,
1996).
Fig. 6.7. Dependence of absorption coefficient a on photon energy for CuInSez: (a) single
crystal, 300 K, (b) single crystal, 100 K, and (c) thin film, 300 K. (Reprinted with permission
from L. L. Kazmerski et al., J . Vac. Sci. Technol. Al, 395 (1983). Copyright 1983,
American Institute of Physics.)
Copper Indium Diselenide and Other I-III-VI Materials 191
Crystal Structure
CuInSez crystallizes in a diamondlike lattice with a face-centered tetragonal
unit cell (Parkes et al., 1973). The lattice structure and unit cell for the
chalcopyrite structure are shown in Fig. 6.8. Each selenium atom serves as the
center of a tetrahedron of two Cu and two In atoms, and each metallic atom
is surrounded by a tetrahedron of selenium atoms. The lattice parameters
are a = 0.5784 nm, and c = 1.1614 nm. CuInSz and CuInTez have similar
structures.
Crystal Defects
The electrical conductivity of CuInSez is apparently largely determined by na-
tive defects with a high degree of compensation between defect donors and
acceptors, and can therefore be varied by the ratio Cu/In used in material fab-
rication by vacuum evaporation and by the pressure of Se used during growth
or annealing.
Neumann (1983) considered vacancies, interstitials and antisite defects in
the cation lattice as possible sources of electrically active defects, and predicted
192 Photovoltaic Materials
that the active defects in p t y p e material are an indium vacancy (K,) and a Cu
antisite defect (CUI,,). In order to achieve low-resistivity, p t y p e CuInSe2, the
Cu/In ratio must be slightly larger than required for stoichiometric material.
Photoluminescence (Abou-Elfotouh et al., 1984) and cathodoluminescence
data (Masse and Redjai, 1984) were interpreted to indicate the following energy
levels for various defects; in Cu-rich material: acceptor CUI, (30 meV above
the valence band edge, V B ) ,acceptor V,, (85 meV above VB) and acceptor VS,
(130 meV above V B ) ,and in In-rich material: acceptor VC, (45 meV above
V B ) ,acceptor Vs, (130 meV above V B ) and donor Inc, (35 meV below the
conduction band edge).
Masse (1990) reviewed the various energy levels of defects with a low energy
of formation, reported for I-111-V12 compounds (e.g. for CuInSe2: Vs,, Vc,,
V,,, and C U ~ , ) ,and concluded that evidence favors a hydrogenic acceptor
associated with a VC, rather than an antisite defect.
Hole and electron traps have been investigated using Deep Level Transient
Spectroscopy (DLTS) in single crystal Schottky junctions based on CuInSe2
(Igalson and Bacewicz, 1994) and in thin film CuInSe2/CdS/ZnO photovoltaic
devices (Igalson and Schock, 1996). The general effect observed is a decrease in
the density of shallow electron traps and an increase in the density of hole traps
after an injection of electrons, metastable below 200 K. It has been proposed
(Igalson, 1993) that this phenomenon can be described appropriately in terms
of a model involving two charge states of the same defect, similar in many ways
to the behavior of the dangling-bond defect in a-Si:H discussed in Chapter 3.
The data suggest that electron traps are converted to hole traps after the
capture of two electrons, and a possible candidate for the defect is a double
donor In/Cu or a complex related to it. Similar indications of metastable defect
behavior in Cu(In,Ga)Sea cells have been reported (Schmitt et al., 1996; Walter
et al., 1996a), and in CuIn(SSe)z films (Herberholz et al., 1996).
A method to deduce energy distributions of defects in the band gap by
measuring the complex admittance of a junction has been applied to hetero-
junctions of pCu(In,Ga)Sea thin films with n-ZnO (Walter et al., 1996b). A
distribution of hole traps with maximum density corresponding to an ioniza-
tion energy of about 0.3 eV was indicated, similar to the results of modulated
photoconductivity (Herberholz et al., 1994) and DLTS measurements (Igalson
and Bacewicz, 1992).
0.5 pm from the photoelectromagnetic effect (Mora and Romeo, 1977), 2.5 pm
from EBIC data on pCuInSe2/CdS heterojunctions (Piekoszewski et al., 1980),
and 1.0 pm from surface photovoltage measured with pCuInSe2/electrolyte
diodes (Johnson et al., 1985), have been reported.
Optical Absorption
Figure 6.7 shows the high values of absorption constant achieved with CuInSe2
which correlate with the high values of J,, found in solar cells based on
CuInSez. Figure 6.9 shows the variation of absorption with photon energy
for single crystal and thin film CuInSe2, indicative of a direct band gap with
magnitude EG for which it would be expected that (cuhv)2 would be propor-
tional to (hv- EG). Differences between the single crystal and thin film curves
in Fig. 6.9 are probably due to small differences in composition.
Fig. 6.9. Dependence of optical absorption of CuInSe2 on photon energy plotted as (ahv)2
versus photon energy hv, which yields a linear relationship for a direct band gap. (m) single
crystal at 300 K, ( 0 ) thin film at 300 K, ( 0 ) thin film at 100 K, and (A) annealed thin film
at 300 K. (Reprinted with permission from L. L. Kazmerski et al., J. Vac. Sci. Technol.
A l , 395 (1983). Copyright 1983, American Institute of Physics.)
194 Photovoltaic Materials
Interface Properties
In Sec. 5.4 the importance of interface interactions in the CdS/CdTe het-
erojunction structure leading to the formation of CdTel-,S, were described.
Similar investigation has been made of the possible reactions between different
species a t the CdS/CuInSez interface. Analyses involving x-ray photon spec-
troscopy (XPS) (Kazmerski et al., 1982) and electron energy loss spectroscopy
(EELS) (Jamjoum et al., 1982) both indicate the presence of a chemical reac-
tion at the CdS-CuInSez interface with the formation of CuzS and Cu2Se as
the reaction products.
Oxidation of CuInSe2
Oxidation studies have been made both on p t y p e polycrystalline thin films fab-
ricated by three-source deposition, and on Bridgman grown crystals (Kazmer-
ski et al., 1981, 1983a). For single crystals oxidized at 140°C XPS shows
the formation of Cu,Se at the dry oxide/CuInSe2 interface. The native ox-
ide formed on a (110) cleaved CuInSez surface by simple exposure to dry
air at room temperature is primarily In203 with less than 5% SeO2. If the
oxidation is thermally induced using either dry oxygen or dry air, the frac-
tion of SeO2 is approximately doubled, and evidence for Cu,Se is seen a t the
interface. For anodic oxidation (exposure to electrolyte solution of sodium
tartrate/HzO/tartaric acid, pH = 6.0 a t room temperature), the fraction of
In203 is reduced and evidence is found for about 10% CuzO, 10% CuO, and
2% Se are found, with CuO at the interface. For CdS/CuInSez cells, a heat
treatment at 200-220°C in oxygen or air consistently improves the operating
characteristics, as illustrated in Fig. 6.10, while heating in high vacuum, Ar,
N, H, or He does not seem to degrade the cell performance.
Non-ideal current-voltage characteristics of thin film CuInSe2 solar cells
such as cross-over of the dark and light current-voltage curves, or strongly
bias dependent current collections, have been attributed to surface oxidation
occurring during several processing steps (Walter et al., 1995). These include
Copper I n d i u m Diselenide and Other I-111-VI Materials 195
I I I I
a2 0.4
Voltage IVI
Fig. 6.10. Effect of heat-treatment in oxygen on light and dark current-voltage characteristics
of CdS/CuInSeZ heterojunction s6lar cell. Dashed line, before 0 2 anneal; solid line, after 0 2
anneal. (Reprinted with permission from R. A. Mickelsen et al., “Cadmium Sulfide/Copper
Ternary Heterojunction Research”, final Rep. XJ-9-8021-1, Boeing Aerospace Co., Seattle,
WA (1982). Copyright 1982, Boeing Aerospace.)
the use of oxygen during the rf sputtering of the ZnO window layer, the air
exposure of the absorber prior to the deposition of the window, and the use of
an In,(OH,S), buffer layer.
Thermal Stability
Low-temperature annealing of the CdS/CuInSez solar cell does not alter its
electrical or interface composition (Kazmerski et al., 1982; Kazmerski, 1983;
Mickelsen and Chen, 1983; Hsiao et al., 1983). Annealing a t 60°C for 7500
hours causes no changes in cell performance or in the compositional appearance
of the interface. Changes do begin to occur when the temperature is raised
above 200°C. Two hours of annealing in Ar at 220°C caused decreases in &,
of 5%, J,, of 17%, f f of 21%, and efficiency of 38%; these results suggest that
the heat treatment has caused growth of the interfacial transition layer, and
XPS identifies the existence of Cu,S and Cu,Se at this interface (Wager et al.,
1983). Heat treatment above 400°C causes a catastrophic degradation of solar
196 Photovoltaic Materials
cell performance due to the rapid diffusion of Cd across the interface region,
an effect that is more severe in polycrystalline thin-film cells than in single
crystal cells, suggesting that grain-boundary diffusion is involved.
Contacts
A basic characteristic of almost all CuInSe2 solar cells is the Mo back con-
tact. Often high-efficiency solar cells show a non-ohmic contact behavior that
has been attributed to the Mo/CuInSea back contact. Earlier measurements
indicated that the nature of the Mo contact to single crystal CuInSez de-
pended on the defect density in the CuInSez (Abou-Elfotouh et al., 1989).
The current-voltage characteristics of the Mo/CuInSez contact on operating
solar cells prepared by selenization of Cu/In precursors in H2Se with the Mo
deposited by sputtering, the CdS by chemical bath deposition, and the ZnO by
rf sputtering, were measured by a technique that allowed separate evaluation
of the contact properties and showed the Mo contact to be completely ohmic
(Shafarman and Phillips, 1996). Simulations suggest that the CdS/CuInSez
interface is the most likely origin of the non-linear current-voltage effects.
Vacuum Evaporation
Three variations of vacuum evaporation have been used for the deposition
of CuInSez films. (1) Single-source deposition or flash-evaporation uses a
Copper I n d i u m Diselenide and Other I-Ill-VI Materials 197
High efficiency solar cells could not be made on CuInSe2 films formed in
a tubular reactor using only HzSe as the reacting gas, but by mixing 0 2 with
the HzSe, CuInSeZ was formed from which cells with efficiencies over 10% were
fabricated (Verma et al., 1993). For OZ/HzSe ratios of 0.9 to 0.2, Cu7Se4 and
In203 were formed along with the CuInSez. But for low 02/H2Se ratios, 0.02
to 0.004, single phase CuInSez films were formed suitable for photovoltaic cells.
It was suggested that the presence of oxygen in a dilute HzSe mixture stabilizes
the morphology of Cu-In films when heat treated at 400°C in Ar or Ar/H2,
controls the amount of Se, species formed in the gas phase, and causes In loss
during CuInSe2 film growth resulting in the formation of a copper selenide
phase.
Yamanaka et al. (1993) evaluated the process of selenization of Cu-In layers
with elemental Se at temperatures between 150" and 650°C. At temperatures
above 250°C it was deduced that In in the liquid phase reacts with Se to form
InSe, then reacts with CuTSe4 to form CuInSe2, which a t these temperatures
is the dominant phase in the film.
A model describing the chemical kinetics of the selenization of Cu-In layers
has been developed (Russell et al., 1994a,b)for both the tubular reactor growth
with HzSe and for a PVD selenization reactor.
A chemical modification of the Mo-coated substrate has been proposed as
a means of obtaining better quality CuInSez films (Basol et al., 1991). Use of
a Te interlayer on the Mo surface before the deposition of C u / h precursors
and subsequent selenization is reported t o produce stoichiometrically, morpho-
logically and mechanically superior CuInSe2 films. It is suggested that thin
interfacial layers of metals such as Cr and T i might also effectively improve
the adhesion of the CuInSe2 films.
An alternate sequential deposition process involves the deposition of InSe,
and Cu at a low substrate temperature followed by a high temperature sel-
enization step, resulting in an efficiency of 14% for solar cells based on these
layers (Zweigart et al., 1995, 1996).
MOCVD
An extensive investigation has been made of the use of MOCVD techniques
to deposit CuInSez thin films for solar cell applications (Sagnes et al., 1992;
Artaud et al., 1994a,b;Ouchen et al., 1995a,b). It was shown that the MOCVD
process performed in a horizontal reactor operating at atmospheric pressure,
with H2 as the carrier gas, was suitable to grow polycrystalline CuInSe2 thin
films with properties comparable to those deposited by vacuum evaporation.
Copper I n d i u m Diselenide and Other I-111-VI Materials 199
The original approach involved a single step codeposition process from three
precursors: TMIn (trimethylindium), (hfa)2 Cu NHEt2 (hexafluoroacetyl-
acetonato copper mixed with diethylamine), and H2Se. This produced ma-
terial suitable for photovoltaic conversion, but with a n efficiency below 3%.
Better results involving enhanced grain size were indicated by a multistep
deposition process using different substrate temperatures, and based on the
growth of a ternary CuInSe2 film on a starting layer of Cu(Se) or In(Se) to
provide quasi-epitaxial growth.
Sputtering
Attempts to deposit CuInSe2 films by radio frequency or magnetron sputter-
ing, a technique that is promising for large-area cell production, result in films
with grain size and reproducibility problems. Samaan et al. (1983) used rf
sputtering in argon gas using p t y p e polycrystalline CuInSe2 ingots or loose
powders as the targets, Piekoszewski et al. (1980) used water-cooled, pressed
powder targets of CuInSe2, and Thornton et al. (1984) used reactive mag-
netron sputtering by simultaneously sputtering Cu and In in an Ar +
HzSe
atmosphere to produce cells with an efficiency of 556%.
In an effort to reduce the density of defects in sputtered CuInSez films,
presumably caused by bombardment by high-energy particles from the Se tar-
get, a technique known as “hybrid sputtering” has been developed in the effort
to make the voltage applied to all targets as low as possible (Nakada et al.,
1995). In this technique, the Se target source of high-energy ions is replaced
by a Se effusion cell. Films deposited by this hybrid sputtering method show
relatively large grain size and low density of defects, enabling the fabrication
of an 11.3% efficient cell.
Spray Pyrolysis
The convenient, low-cost method of spray pyrolysis has been used t o produce
thin-film CuInSe2 by spraying a freshly prepared aqueous solution of saturated
CuC1, InC13 and N-dimethyl selenide in the ratio 4:1:0.5 onto a heated substrate
(Pamplin and Feigelson, 1979; Gorska et al., 1980). Mooney et al. (1980)
used spray pyrolysis to produce low-resistivity n-type CuInSe2 which was then
converted to p t y p e by heat-treatment in H2/H2Se at 6OO0C, and showed a
photovoltaic response in sprayed CdS/CuInSez junctions. The technique has
not yet been shown to be suitable for general solar cell applications.
200 Photovoltaic Materials
Electrodeposition
Pern et al. (1987) described the preparation of CuInSea films by a one-step
electrodeposition process. The film quality was affected by the concentration
of the buffering agent ethylenediamine, which must be kept below 0.125 M to
produce photovoltaically useful films.
The phases obtained under various conditions of potential and flux ratio in
the application of electrodeposition techniques t o the preparation of CuInSe2
films have been investigated (Thouin and Vedel, 1995; Vedel et al., 1996).
It has been concluded that a combination of one-step electrodeposition and
annealing treatment in a Se atmosphere is a promising approach for improv-
ing the quality of electrodeposited CuInSez films for photovoltaic applications.
(Guillemoles et al., 1996; Lincot et al., 1994). The properties of electrode-
posited/selenized films were investigated by luminescence measurements, Hall
effect and photoelectrochemical characterization. Films are p t y p e , with carrier
densities of 10'6-10'7 cmP3, and diffusion length of about 1 pm. An efficiency
of 6.5% is reported for the best CuInSea/CdS/ZnO solar cell prepared in this
way to date. Interface recombination, not bulk properties, is shown to be the
dominant loss mechanism.
1993, 1994). The deposits were made on glass or Mo substrates below 620°C in
glass tubes sealed under vacuum after the introduction of solid iodine. Films
obtained above 550°C have a composition close to the stoichiometric compo-
sition. Except for low iodine concentrations (< 0.1 mg cmP3) or high iodine
concentrations (> 0.5 mg ~ m - ~ film
) , growth transport is governed by diffu-
sion.
Films deposited on SnO2 show a new photovoltaic effect with the positive
polarity on the SnO2, presumably because of the formation of a doping gradient
in the CuInSe2 films due to the deposition conditions (Masse and Djessas,
1995).
by two layers, a lower undoped layer next to the CuInSez and an upper n-type
doped layer. Heat treatment of the deposited cell for 10-40 minutes in oxygen
at 225"C, needed to achieve optimum efficiency, apparently provides a grada-
tion in composition of the two CuInSe2 layers. Cells with areas up to 91 cm2
were prepared with an efficiency of 9.5%. These cells had the following struc-
ture: soda lime glass/2-3 pm Mo/3 pm CuInSe2/2 pm ZnCdS/Al-/SiN/SiO.
Early anti-reflection coatings of MgFZ/SiO, were replaced by plasma enhanced
chemical vapor deposition at 175°C of amorphous hydrogenated SIN and SiO
films.
Several investigations were made of the effect of the specific composition of
the CuInSe2. Noufi et al. (1984) used the three-source deposition method to
prepare a large number of films, varying the Cu/In ratio from 0.4 to 1.2 from
film to film, while keeping the In and Se rates constant. The results indicated
that the control of the electrical properties of CuInSea films, and hence the
performance of solar cells using them, required good control of stoichiometry
since the electrical properties are dominated by native defects.
Noufi and Dick (1985) examined the compositional and electrical proper-
ties of the two-layer CuInSea structure characteristic of the cells prepared by
Mickelsen et al. (1987). They found that in the final solar cell device, the
two individual CuInSe2 layers had mixed and remained p t y p e during the cell
processing steps, except for the very last few tenths of a micrometer to deposit,
which was Cu-deficient and of high resistivity; the two CdS layers consisted of
an almost stoichiometric layer close to the junction and a top In-doped low-
resistivity layer. Rocheleau e t al. (1987) investigated the tolerance of CuInSeZ
solar cells of this type to variations in film composition. They found that cell
efficiencies of 10% or better could be obtained for a range of Cu from 23% to
27% and a range of indium from 25% to 28%.
The effects and mechanisms involved in the required air-bake to optimize
these cells has been investigated. Noufi et al. (1985, 1986a) have studied the
effect. of the air-bake on the spectral response of CdS/CuInSeZ cells, alternate
methods of oxidizing or reducing the CuInSe2 (Noufi et al., 1986b), the depen-
dence of the electrical conductivity of CuInSe2 on composition and oxidation
(Datta et al., 1985), and the variation of the thermally stimulated conductivity
of CuInSe2 with oxidation (Datta et al., 1986). Their results can be summa-
rized as follows: (1) improvements in cell performance on air-baking consist
primarily of increases in open-circuit voltage and fill factor; (2) air-baking was
beneficial only if the temperature was not greater than 2OO0C, degradation
being caused by higher temperatures; (3) the effect of the air-bake was to
Copper I n d i u m Diselenide and Other I-III-VI Materials 203
introduce oxygen into the CuInSez with effects that could be duplicated by the
use of chemical oxidants or reversed by the use of chemical reducing agents such
as hydrazine or by exposure t o an electron beam; (4)the effect of oxygen in the
CuInSe2 films was to increase the p t y p e electrical conductivity, presumably
by the compensation of donors associated with interstitial In or In-on-Cu-site
defects, as suggested also by a decrease in the thermally stimulated conductiv-
ity associated with these levels, or by tying up dangling bonds at the surface
of selenium-deficient CIS grains. Damaskinos et al. (1987) showed that the re-
sistivity of the CuInSe2, which depends on its state of oxidation, is responsible
in the solar cell for large reverse saturation current, low open-circuit voltage,
and large cross-over between dark and light current versus voltage curves. All
these phenomena appeared to be reversible by oxidation/reduction cycles.
Substrates
The commonest substrate for the CuInSez solar cell is Mo on soda-lime glass.
Effects due to Na diffusion from this glass substrate into the CuInSe2 layer
during cell fabrication by HzSe selenization have been investigated (Basol et al.,
1994). Excessive Na diffusion through the Mo layer and its reaction with the
processing atmosphere at the Mo/CuInSe2 interface may limit the efficiency
due to the formation of a high-resistivity layer near this contact. On the other
hand, a moderate level of Na diffusion through the Mo layer dopes the CuInSe2
layer p t y p e and improves the efficiency of the cells, especially those with low
Cu/In ratios.
The influence of substrates, particularly alkali free substrates versus soda
lime glass, on Cu(In,Ga)Se2 thin films has been investigated by Ruckh et al.
(199613). The presence of Na in films on soda lime glass is correlated with
an enhanced formation of S e - 0 , In-0, and Ga-0 bonds a t the surface after
several days of exposure to air, the electrical conductivity is one order of mag-
nitude higher, and solar cells prepared on these substrates exhibit increased
open-circuit voltage. Junctions prepared on alkali free substrates show an in-
creased space charge width. These results can be explained in terms of a higher
acceptor density in the films prepared on soda lime glass.
Efforts have been directed toward developing flexible substrates for CuInSez
solar cells. The first efficient CuInSez cell on a flexible metal substrate was
described by Basol et al. (1993). A selenization technique was used involving
H2Se and e-beam evaporated Cu/In precursor layers t o obtain CuInSez. Mo,
Ti, and A1 foils were evaluated as possible flexible substrate materials for solar
cells. Flexible cells (capable of being bent to a radius of curvature of about
2 cm) upon Mo foil substrates had a conversion efficiency of above 8%.
Problems associated with these metal films when large-scale module fab-
rication was considered, led to the investigation of an insulating, light-weight
and flexible polymeric substrate: a 50 pm thick KAPTON polyimide sheet
supplied by Dupont (Basol et al., 1996). The best 1 cm2 area device showed a
total area efficiency of 8.7%.
1.6
1.5
1.4
1.3
1.2
1.I
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Ga/( I n+Ga)
(4
0.9
0.8
0.7
0.6
0.2 0.3 0 . 4 0.5 0.6 0.7 0.8 0.9
G a/( I n + G a )
(b)
Fig. 6.11. Variation of properties of Cu(1nGa)Sez solar cells deposited by four-source vacuum
evaporation as a function of the Ga/(In + Ga) ratio. (a) Band gap, (b) open-circuit voltage,
(c) short-circuit current, mA/cm2, and (d) solar cell efficiency. (Data from Shafarman et al.,
1996.)
Copper Indium Diselenide and Other 1-Ill-VI Materials 209
Ga/( In+Ga)
(c)
Ga/( I n+ G a )
(4
Fig. 6.11. (Continued )
have been deposited in thin film form over the range of 0 < x < 1 by coevapo-
ration (Contreras et al., 1996a,b). The band gap of these defect chalcopyrites
is about 0.20 eV larger than an equivalent Cu(In,Ga)Sez material with similar
Ga content. All defect materials of this type prepared to date show very high
resistivity, low mobility, and low carrier concentration, but do seem to show a
change in conductivity type from n-type to p t y p e as Ga is increased beyond
x > 0.3.
A related investigation of secondary phases in the surface of coevaporated
Cu-In-S films shows that In-rich films have an In-rich second phase at both
surfaces, which has the cation ratio of CuIn3S5 (Scheer and Lewerenz, 1995).
The results are interpreted as the formation of a phase of CuIn3S5 at the grain
boundaries of In-rich CuInS:! films. By way of contrast, Cu excess in Cu-In-S
films leads to a segregation of CuS at the front surface of the films, and of
MoS2 at the back Mo/CuInSz interface.
Controlled incorporation of Na and S into the absorber layer of Cu(In,Ga)
Sez solar cells has been investigated. Cu(In,Ga)Sez solar cells are reported
to show appreciable improvement in performance when Na is present at the
growth surface of films deposited by coevaporation (Rockett et al. 1996). Al-
though it is reported that Na segregates rapidly out of the Cu(In,Ga)Sez for
these deposition conditions, it leads to a strong (112) preferred orientation of
the Cu(In,Ga)Sez and apparently improves the cell performance. In a related
investigation of the effects of Na and S during the rapid thermal compound
formation process (Karg et al., 1993) involving Cu/Ga sputtering and Se evap-
oration, a Na-related shallow acceptor state at about 75 meV above the valence
band was identified, and S incorporation apparently led to an increased open-
circuit voltage because of a reduction in electrically active deep trap states
(Rau et al., 1996).
Nakada et al. (1997) investigated the effects of Na on Cu(1nGa)Sez solar
cells by a multistep process in which NazSe and Cu were evaporated simulta-
neously. The hole density could be held in the 10l6 to 1017 range for
a wide range of Cu/(In +
Ga) ratios of 0.4 to 0.8 by the addition of NazSe,
whereas without the NazSe, the hole density dropped below 10l6 cm-3 for
+
Cu/(In Ga) less than 0.8. Solar cell efficiencies of 10 to 13.5% were achieved
over a wide range of Cu/(In + Ga) ratios from 0.51 to 0.96. It was also found
that p t y p e Cu(In,Ga)3Ses phase films, which co-exist with the Cu(In,Ga)Sez
phase for Cu/(In + Ga) < 0.71, could be obtained for the first time with a
hole density high enough t o allow use as an absorber layer for Cu/(In + Ga)
ratios in the range below 0.51, producing a 12% efficient solar cell for Cu/(In
+ Ga) = 0.51.
Copper I n d i u m Diselenide and Other I-III-VI Materials 211
C(4
120 .
11.8 .
lla .
llr)
112 .
110 *
Fig. 6.12. Dependence of the lattice parameters and the energy gap on the composition z of
CuGa(SezTel-,)2 alloys. (Reprinted with permission from G. Masse et al., J . Appl. Fhys.
74, 1376 (1993). Copyright 1993 American Institute of Physics.)
from the CuGaTe2 phase diagram. Thermal evaporation gave similar results,
but flash evaporation gave thin films with a composition close to that of the
source.
A general multinary system of form C u ~ _ , A g , I n ~ ~ y ~ , G a y A 1 , S e ~ ~ l ~ , ~ ,
S2,TezW has been modeled by Dhingra and Rothwarf (1996), including graded
band gap effects to improve the open-circuit voltage without sacrificing the
short-circuit current in CuInSea/CdS solar cells. The model is based on a pi-
n structure with the graded bandgap material being the intrinsic region. Higher
efficiency appears to be possible as a result of confining the incident light in the
intrinsic region to achieve higher absorption efficiency, grading the conduction
and valence bands to improve electron and hole collection efficiency with op-
timal grading being linear from 1.5 to 1.3 eV, and tailoring the recombination
region to minimize recombination and achieve higher open-circuit voltage.
6.6. M u l t i j u n c t i o n Cells
6.7. General M a t e r i a l s C o n s i d e r a t i o n s
The fabrication and use of CuInSe2 solar cells raises several questions in the
areas of health and safety, and the supply of key elements (Zweibel and Barnett,
1993).
Hazardous chemicals like H2Se and Cd are involved in the fabrication and
structure of these cells. H2Se is a highly toxic gas but it can be used safely
provided that certain well-known safety precautions are used (Moskowitz et al.,
1986;Bottenberg and Sproull, 1988). Risks associated with it can be minimized
if the gas is produced on-site and is recycled after use. Sources of Se are being
sought that would eliminate the need for H2Se completely.
214 Photovoltaic Materials
The toxic dangers associated with Cd are certainly less in the CdS/CuInSez
cells than in the CdS/CdTe cells, for which the issue has been discussed in
Chapter 5, because of the smaller amount of Cd involved.
The use of Se in the cells requires proper disposal techniques to avoid
contamination of groundwater. These appear to be available.
If CuInSe2 cells enter into large-scale production, the availability of In might
become an issue. Although it appears that there is a sufficient abundance of
In to meet the growing demand of an effective solar cell market, there could
be a conflict between supply and demand. Requirements for large amounts of
In could be reduced by using thinner layers of CuInSe;! made possible by the
high absorption constant, using CuGaInSe2 alloys in which about 25% of the
In is replaced by Gal and recycling cells after 30 years.
CHAPTER 7
7.1. Overview
Earlier in this book we have stressed the fact that the actual number of dif-
ferent materials capable of being used as the light-absorbing material in sta-
ble, highly efficient, semiconductor junction, solar cells is strikingly small: Si,
GaAs, CdTe, and CuInSe2, with variations due to material structure and re-
lated alloys, represents the whole list. In this final chapter we take a brief look
at some of the other materials and structures that either have been promising
at some time in the past, or might be promising in the future, which have not
been discussed previously in this book. Since 1993 a summary of the highest
confirmed efficiency for a range of photovoltaic cells has been published every
six months, including new entries as they qualify (Green, 1996).
215
216 Photovoltaic Materials
0 2 4 6
Time (minutes)
8
i10
Fig. 7.1. Data from Auger spectroscopy on a Zn3Pz surface left under the electron beam,
showing a decrease in the magnitude of the P signal with time. (Reprinted with permission
from S. Casey et al., J . A p p l . Phys. 61, 2944 (1987). Copyright 1987 American Institute of
Physics).
I
101~
'-Pi
-/-
I
/'AL20,/Si
al
:
v)
10'0
Fig. 7.2. Calculated surface state density as a function of semiconductor surface potential for
an AI/A1203/Zn3P2 device. Also shown are the results of similar experiments with A1203
electron-beam evaporated on Si (le Contellec and Morin, 1978) and InP (Favennec et al.,
1979), and the results of a similar analysis on an AI/SiOz/SiMOS device (T. Akinwande,
1987). (Reprinted with permission from S. Casey et al., J . A p p l . Phys. 61,2944 (1987).
Copyright 1987 American Institute of Physics).
over most of the range above its band gap energy, like CuInSez), excellent
chemical stability, and the possibility of being prepared in either n- or p t y p e
form. Schottky barrier AllpWSez solar cells have been prepared with an
efficiency of 5.3% (Clemen and Bucher, 1978). Difficulties were encountered
with depositing films with the required structural properties for this layered
material (Miller, 1981).
8
@$...+"...N$
N13N
Fig. 7 . 3 . Molecular structures of some interesting organic materials: (a) copper phthalo-
cyanine (CuPc); (b) perylene tetracarboxylic derivative (PV); (c) hydroxysquarylium; (d)
merocyanine dye. (Reprinted with permission from P. Di Marco and G. Giro in Organic Con-
ductors, p. 791 (1994) by courtesy of Marcel Dekker Inc., Copyright 1994, Marcel Dekker,
NY).
In. Reasonable open-circuit voltages are produced, but only small short-circuit
currents. Solar efficiencies are usually less than 1%.
In p-n junction cells, illumination can be through a transparent electrode
onto the n-type material, chosen to be the shorter-wavelength absorber with
a fluorescence band that overlaps the absorption band of the p t y p e material.
A notable improvement over earlier work was obtained by Tang (1986) using
a cell pictured in Fig. 7.4, composed of ITO/copper-phthalocyanine (CuPc,
ptype)/perylene tetracarboxylic derivative (PV, n-type)/Ag. Under AM2
C o p p e r I n d i u m Diselenide a n d O t h e r I-III-VI Materials 221
I
I
t
Light
Fig. 7.4. Schematic diagram of a p n organic cell between p t y p e CuPc, and n-type PV.
(Reprinted with permission from C. W. Tang, A p p l . Phys. Lett. 48, 183 (1986). Copyright
1986 American Institute of Physics).
225
226 Photovoltaic Materials
A. K. Berry, J. L. Boone and T. P. Van Doren, Materials Letters 10, 261 (1990).
K. A. Bertness, M. Ladle Ristow and H. C. Hamaker, 20th I E E E Photovoltaic Spe-
cialists Conference, IEEE Publishing, NY (1988), p. 769.
K. A. Bertness, S. R. Kurtz. D. J. Friedman, A. E. Kibbler, C. Kramer and J. M.
Olson, 24th IEEE Photovoltaic Specialists Conference, IEEE Publishing, NY
(1994), p. 1671.
J. S. Best, J. 0. McCaldin, T. C. McGill, C. A. Mead and J. B. Mooney, Appl. Phys.
Lett. 29, 433 (1976).
A. Bett, S. Cardona, A. Ehrhardt, F . Lutz, H. Welter and W. Wettling, 22nd I E E E
Photovoltaic Specialists Conference, IEEE Publishing, NY (1991), p. 137.
A. Bett, K. Borgwarth, Ch. Schetter, 0.V. Sulima and W. Wettling, 6th International
Photovoltaic Science and Engineering Conference, New Delhi (1992), p. 843.
M. Bettini, K. J. Bachmann, E. Buehler, J. L. Shay and S. Wagner, J. Appl. Phys.
48, 1603 (1977).
M. Bettini, K. J. Bachmann and J. L. Shay, J. Appl. Phys. 49, 865 (1978).
W. Beyer, H. Me11 and H. Overhof, in Amorphous and Liquid Semiconductors (W.
E. Spear, ed.) CICL, Edinburgh (1977), p. 328.
M. T. Bhatti, K. M. Hynes, R. W . Miles and R. Hill, Int. J . of Solar Energy 12, 171
(1992).
M. Bhushan and A. Catalano, Appl. Phys. Lett. 38, 39 (1981).
M. Bhushan, Appl. Phys. Lett. 40, 51 (1982).
M. Bhushan, Final Report, SERI Subcontract # XE-2-02048-1 (1984).
R. N. Bicknell, N. C. Giles and J. F. Schetzina, Appl. Phys. Lett. 49, 1095, 1735
(1986).
R. N. Bicknell, N . C. Giles, J. F. Schetzina and C. Hitzman, J . Vac. Sci. Technol.
A5,3059 (1987).
R. N. Bicknell-Tassius et al., Appl. Surface Sci. 36,95 (1989).
R. N. Bicknell-Tassius et al., J . Cryst. Growth 101, 33 (1990).
R. W. Birkmire, R. B. Hall and 3. E. Phillips, Proc. 17th I E E E Photovoltaic Special-
zsts Conference, IEEE Publishing, NY (1984), p. 882.
R. W. Birkmire, B. E. McCandless, W. N. Shafarman and R. D. Varrin, Proc. 19th
European Photovoltaic Solar Energy Conference (1989), p. 134.
R. W. Birkmire, B. E. McCandless and S. S. Hegedus, Int. J. Solar Energy 12, 145
(1992b).
R. W. Birkmire, S. S. Hegedus, B. E. McCandless, J. E. Phillips, T. W . F. Russell,
S. N. Shafarman, S. Verma and S. Yamanaka, Photovoltaic Advanced Research
and Development Project, AIP Conference Proceedings 268, 212 (1992).
R. W. Birkmire and P. V. Meyers, First World Conference on Photovoltaic Energy
Conversion, 24th IEEE Photovoltaic Specialists Conference, IEEE Publishing,
NY (1994), p. 76.
R. W. Birkmire, H. Hichri, R. Klenk, M. Marudachalam, B. E. McCandless, J. E.
Phillips, J. M. Schultz and W. N. Shafarman, 13th N R E L Photovoltaic Program
Review Meeting, Lakewood, CO (1995).
U. Blieske, S. Sterk, A. Bett, J. Schumacher, W. Wettling, A. Marti, M. J. Terron
and A. Luque, First World Conference on Photovoltaic Energy Conversion,
230 Photovoltaic Materials
J. Kopp, J. Knobloch and W. Wettling, 11th E.C. Photovoltaic Solar Energy Con-
ference, Montreux, Switzerland, 49 (1992).
P. Kordos, R. A. Powell, W. E. Spicer, G. L. Pearson and M. B. Panish, Appl. Phys.
Lett. 34,366 (1979).
P. Kordos and G. L. Pearson, Solid-state Electronics 23,399 (1980).
A. V. Koval, V. Z. Nikorich, A. V. Simashkevich, R. L. Sobolevskaya and K. D.
Sushkevich, J . Cryst. Growth 110,915 (1991).
G. Kroetz, J. Wind, H. Stitzl, G. Mueller, S. Kalbitzer and H. Mannsperger, Phil
Mag. B 63,101 (1991).
U. Kroll, F. Finger, J. Dutta, H. Keppner, A. Shah, A. Howling, J.-L. Dorier and
Ch. Hollenstein, Mats. Res. SOC.Symp. Proc. 258,135 (1992).
L.-C. Kuo, Y.-T. Tsai, W.-Y. Hsu and N.-H. Lu, First world Conf. on Photovoltaic
Energy Conversion, 24th I E E E Photovoltaic Specialists Conference, IEEE Pub-
lishing, NY (1994), p. 690.
H. Kurita, T. Takamoto, E. Ikeda and M. Ohmori, 7th International Conference on
Indium Phosphide and Related Materials (1995), p. 516.
S. R. Kurtz, J. M. Olson and A. Kibbler, 2 l s t I E E E Photovoltaic Specialists Confer-
ence, IEEE Publishing, NY (1990), p. 138.
W. Kusian, E. Guenzel and R. D. Plaettner, Solar Energy Materials 23,303 (1991).
W.Kusian, K.-D. Ufert and H. Pfleiderer, Solid State Phenomena 44-46,823 (1995).
A. Kylner, J. Lindgren and L. Stolt, J . Electrochem. SOC.143,2662 (1996).
A. Kylner and M. Wirde, Jpn. J. Appl. Phys. 36,2167 (1997a).
A. Kylner and E. Niemi, 14th European Photovoltaic Solar Energy Conference, Spain,
June-July (1997b).
Y. Kuwano, S. Tsuda, N. Nakamura, M. Nishikuni, K. Yoshida, T. Takahama,
M. Isomura and S. Nakano, AZP Conf. Proc. 157,126 (1987).
G. Langguth, M. Schoefthaler, T. Sameshima and J. H. Werner, 13th European Pho-
tovoltaic Solar Energy Conference, Nice, (1995).
P. G . LeComber and W. E. Spear, in Amorphous Semiconductors, 36,of Topics in
Applied Physics (M. H. Brodsky, ed.) Springer, NY (1986), p. 251.
C.-T. Lee and R. H. Bube, J. Appl. Phys. 54,7041 (1983).
C.-T. Lee and R. H. Bube, J. Appl. Phys. 5 8 , 880 (1985).
X. Y . Lee, A. K. Verma, C. Q. Wu, M. Goertemiller, E. Yablonovitch, J. Eldredge and
D. Lillington, 25th ZEEE Photovoltaic Specialists Conference, IEEE Publishing,
NY (1996), p. 53.
C. Leguijt, P. Loelgen, A. R. Burgers, J. A. Eikelboom, R. A. Steeman, W. C. Sinke,
P. F. A. Alkemade and P. M. Sarro, J. Appl. Phys. 78,6596 (1995).
D. H. Levi, H. R. Moutinho, F. A. Hasoon, B. M. Keyes, R. K. Ahrenkiel, M. Al-
Jassim, L. L. Kazmerski and R. W. Birkmire, First world Conference on Pho-
tovoltaic Energy Conversion, 24th ZEEE Photovoltaic Specialists Conference,
IEEE Publishing, NY (1994), p. 127.
J. D. Levine, G. B. Hotchkiss and M. D. Hammerbacher, 22nd I E E E Photovoltaic
Specialists Conference, IEEE Publishing, NY (1991), p. 1045.
J. D. Levine, 1992 Photovoltaic Advanced Research and Development Project Con-
ference, Denver, American Institute of Physics Press, NY (1992), p. 46.
References 245
p. 1219.
H. M. Manasevit, K. L. Hess, P. D. Dapkus, R. P. Ruth, J. 3. Yang, A. G. Campbell,
R. E. Johnson, L. A. Moudy, R. H. Bube, L. B. Fabick, A. L. Fahrenbruch and
M.-J. Tsai, 13th IEEE Photovoltaic Specialists Conference, IEEE Publishing,
NY (1978), p. 165.
J. Mandelkorn, C. McAfee, J. Kesperis, L. Schwartz and W. Pharo, J. Electrochem.
Soc., 313 (1962).
D. Mao, L. H. Feng, Y. Zhu, J. Tang, W. Song, R. Collins, D. L. Williamson and
J . U. Trefny, 13th NREL Photovoltaics Program Review, A I P Conference Pro-
ceedzngs 353, 352 (1995).
M. Marudachalam, R. W. Birkmire, J. M. Schultz and T. Yokimcus, First World
Conference on Photovoltaic Energy, 24th IEEE Photovoltaic Specialists Con-
ference, IEEE Publishing, NY (1994), p. 234.
E. Maruyama, A. Terakawa, K. Sayama, K. Ninomiya, Y. Hishikawa, H. Tarui, S.
Tsuda, S. Nakano and Y. Kuwano, 23rd IEEE Photovoltaic Specialists Confer-
ence, Louisville, IEEE Publishing, NY (1993a).
E. Maruyama, Y. Hoshimine, A. Terakawa, K. Sayama, K. Ninomiya, Y. Hishikawa,
H. Tarui, S. Tsuda, S. Nakano and Y. Kuwano, Materials Research Society,
San Francisco (199313).
E. Maruyama, S. Tsuda and S. Nakano, Solid State Phenomena 44-46, 863 Scitec
Publications (1995).
E. Maruyama, Y. Hishikawa, M. Tanaka, S. Kiyama and S. Tsuda, Materials Research
Society, San Francisco (1996).
G. Masse and E. Redjai, J . Appl. Phys. 56, 1154 (1984).
G. Masse, J. Appl. Phys. 68, 2206 (1990).
G. Masse and K. Djessas, Thin Solid Films 226, 254 (1993).
G. Masse and K. Djessas, Thin Solid Films 237, 129 (1994).
G. Masse and K. Djessas, Thin Solid Films 257, 137 (1995).
G. Masse, K. Djessas and L. Yarzhou, J. Appl. Phys. 74, 1376 (1993).
J. H. Matlock, Semiconductor International 2, 33 (1979).
A. Matsuda, S. Yamasaki, K. Nakagawa, H. Okushi, K. Tanaka, S. Iizima, M. Mat-
sumura and H. Yamanioto, J p n . J. Appl. Phys. 19, L305 (1980).
H. Matsumoto, K. Kuribayashi, H. Uda, Y. Komatou, A. Nakano and S. Ikegami,
Solar Cells 11, 367 (1984).
R. H. Mauch, J. Hedstrom, D. Lincot, M. Ruckh, J. Kessler, R. Klinger, L. Stolt,
J. Vedel and H. Schock, Proc. 22nd IEEE Photovoltaic Specialists Conference,
IEEE Publishing, NY (1991a).
R. H. Mauch, M. Ruckh, J. Hedstrom, D. Lincot, J. Kessler, R. Klinger, L. Stolt,
J. Vedel and H. W. Schock, 10th European Photovoltaic Solar Energy Confer-
ence (1991b), p. 1415.
B. E. McCandless, R. W. Birkmire, W. A. Buchanan, J. E. Phillips and R. E. Roche-
leau, 20th IEEE Photovoltaic Specialists Conference, IEEE Publishing, NY
(1988), p. 381.
B. E. McCandless and S. S. Hegedus, 22nd IEEE Photovoltaic Specialists Conference,
IEEE Publishing, NY (1991).
248 Photovoltaic Materials
Listed Chronologically
268
Bibliography 269
Listed Chronologically
270
Review Publications 271
E. Bucher, “New Technologies and Materials for Photovoltaic Solar Energy Conver-
sion”, Electrochemical Society Extended Abstracts 84-2,Electrochemical Soci-
ety, Pennington, NJ (1984), p. 450.
D. E. Carlson, “(a-Si:H) Solar Cells”, in Hydrogenated Amorphous Silicon 21 A, B,
C and D, Semiconductors and Semimetals (J. I. Pankove, ed.) Academic Press,
Inc. (1984), p. 7.
S. J. Fonash and A. Rothwarf, “Heterojunction Solar Cells”, in Current Topics in
Photovoltaics (T. J. Coutts and J . D. Meakin, eds.) Academic Press, London
(1985), p. 2.
L. M. Fraas, “Advanced Concentrator Solar Cells”, in Current Topics i n Photovoltaics
(T. J. Coutts and J. D. Meakin, eds.) Academic Press, London (1985), p. 170.
Y. Hamakawa, “Amorphous-Silicon Solar Cells”, in Current Topics in Photovoltaics
(T. J. Coutts and J . D. Meakin, eds.) Academic Press, London (1985), p. 111.
R. Hill and J. D. Meakin, “Cadmium Sulphide-Copper Sulphide Solar Cells”, in Cur-
rent Topics i n Photovoltaics (T. J. Coutts and J. D. Meakin, eds.) Academic
Press, London (1985), p. 223.
G. Hodes, S. J. Fonash, A. Heller and B. Miller, “Photoelectrochemical Cells Based
on Polycrystalline Semiconductors”, Advances in Electrochemistry and Elec-
trochemical Engineering 13, 113 (1985).
L. Kazmerski and S. Wagner, “Cu-Ternary Chalcopyrite Solar Cells”, in Current
Topics i n Photovoltaics (T. J. Coutts and J. D. Meakin, eds.) Academic Press,
London (1985), p. 41.
J. D. Meakin, “Status of CuInSez Solar Cells”, Photovoltaics, SPIE Vol. 543 Society
of Photo-Optical Instrumentation Engineers, Bellingham, WA (198S), p. 108.
K. Rajeshwar, “Materials Aspects of Photoelectrochemical Energy Conversion”,
Journal of Applied Electrochemistry 15,1 (1985).
T. J. Coutts, L. L. Kazmerski and S. Wagner, eds., Copper Indium Diselenide for
Photovoltaic Applications, Elsevier, NY (1986).
R. H. Bube, “Thin-Film Polycrystalline Solar Cells”, ISES Solar World Congress,
Hamburg (1987).
A. L. Fahrenbruch, “Ohmic Contacts and Doping of CdTe”, Solar Cells 21, 399
(1987).
Y. Hamakawa, “Photovoltaic Power”, Scientific American 256, 76 (1987).
A. Rothwarf, “Criteria for the Design of High Efficiency Thin Film Solar Cells: Theory
and Practice”, Solar Cells 21, 1 (1987).
B. M. Basol, “Electrodeposited CdTe and HgCdTe Solar Cells”, Solar Cells 23, 69
(1988).
E. C. Boes, “A Summary of Recent Photovoltaic Concentrator Technology Develop-
ments”, 20th I E E E Photovoltaic Specialists Conference, IEEE Publishing, NY
(1988), p. 21.
R. H. Bube, “CdTe Junction Phenomena”, Solar Cells 23,1 (1988).
T. L. Chu, “Cadmium Telluride Solar Cells”, Curr. Top. Phovotaics 3, 236 (1988).
H.Curtins and M. Favre, “Surface and Bulk States Determined by Photothermal De-
flection Spectroscopy”, Amorphous Silicon and Related Materials (H. Fritzsche,
ed.) World Scientific Publishing (1988), p. 329.
272 Photovoltaic Materials
H. Takamura, M. Ura and M. Ohmori, “InP Solar Cells and their Flight Ex-
periments”, The Institute of Space and Astronautical Science, Report No. 644,
Kanagawa, Japan (1992).
B. M. Basol, “Preparation Techniques for Thin Film Solar Cell Materials: Processing
Perspective”, Jpn. J. Appl. Phys. 32, Suppl. 32, 35 (1993).
E. C. Boes and A. Luque, “Photovoltaic Concentrator Technology”, in Renewable
Energy: Sources for Fuels and Electricity (T. B. Johansson, H. Kelly, A. K. N.
Reddy, and R. H. Williams, eds.; L. Burnham, ex.ed.) Island Press, Washing-
ton, D.C. (1993), p. 361.
R. H. Bube and K. W . Mitchell, “Research Opportunities in Polycrystalline Com-
pound Solar Cells”, J . Electron. Mat. 22, 17 (1993).
R. H. Bube, LLSolarCells”, in Handbook on Semiconductors, Vol. 4 , Device Physics
(C. Hilsum, ed.) Elsevier, Amsterdam (1981); Revised edition (1993).
D. E. Carlson and S. Wagner, “Amorphous Silicon Photovoltaic Systems”, in Re-
newable Energy: Sources f o r Fuels and Electricity (T. B. Johansson, H. Kelly,
A. K. N. Reddy and R. H. Williams, eds.; L. Burnham, ex.ed.) Island Press,
Washington, D.C. (1993), p. 403.
M. A. Green, “Silicon Solar Cells: Evolution, High-Efficiency Design and Efficiency
Enhancements”, Semicond. Sci. Technol. 8, 1 (1993a).
M. A. Green, ‘Crystalline- and Polycrystalline-Silicon Solar Cells”, in Renewable
Energy: Sources for Fuels and Electricity (T. B. Johansson, H. Kelly, A. K. N.
Reddy and R. H. Williams, eds.; L. Burnham, ex.ed.) Island Press, Washington,
D.C. (1993b), p. 337.
H. Kelly, “Introduction to Photovoltaic Technology”, in Renewable Energy: Sources
for Fuels and Electricity (T. B. Johansson, H. Kelly, A. K. N. Reddy and R.
H. Williams, eds.; L. Burnham, ex.ed.) Island Press, Washington, D.C. (1993),
p. 297.
W. Luft, H. M. Branz, V. L. Dalal, S. S. Hegedus and E. A. Schiff, “Recent Progress in
Amorphous Silicon PV Technology,” N R E L 12th PV Program Review, Denver
(1993).
W. Paul, R. A. Street and S. Wagner, “Hydrogenated Amorphous Semiconductors” ,
J . Electronic Materials 22, 39 (1993).
A. Zunger, S. Wagner and P. M. Petroff, “New Materials and Structures for Photo-
voltaics”, J. Electronic Materials 22, 3 (1993).
K. Zweibel and A. M. Barnett, “Polycrystalline Thin-Film Photovoltaics” , in Re-
newable Energy: Sources for Fuels and Electricity (T. B. Johansson, H. Kelly,
A. K. N. Reddy and R. H. Williams, eds.: L. Burnham, ex.ed.) Island Press,
Washington, D.C. (1993), p. 437.
R. R. Arya, R. S. Oswald, Y. M. Li, N. Maley, K. Jansen, L. Yang, L. F. Chen,
F. Willing, M. S. Bennett, J. Morris and D. E. Carlson, “Progress in Amor-
phous Silicon Based Multijunction Modules”, First World Conference o n So-
lar Energy Conversion, 24th I E E E Photovoltaic Specialists Conference, IEEE
Publishing, NY (1994), p. 394.
P. A. Basore and J. M. Gee, “Crystalline-Silicon Photovoltaics: Necessary and Suf-
ficient”, I E E E First World Conference o n Photovoltaic Energy Conversion,
274 Photovoltaic Materials
277
278 Index
solar spectrum, 5
air mass. 5